PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.
This PDF file contains the front matter associated with SPIE Proceedings Volume 11928, including the Title Page, Copyright information, and Table of Contents.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Aiming at the problem that the complex environment of the paddy field affects the harvesting navigation path extraction
during the visual navigation of the combine harvester, and also to improve the efficiency of the navigation path
extraction, this paper proposes a navigation path extraction method based on the identification of unharvested areas. First,
the Cb color channel of the image is extracted, and the image is initially segmented by adaptive threshold segmentation;
then, the detection of the connected domain area and the unharvested area of the paddy field operation model are used to
achieve the accurate segmentation of the unharvested area, and the operation boundary line pixel coordinate constraints
are added to improve the probabilistic Hough transform algorithm by combining the characteristics of the combine
harvester operation, and the boundary line slope is used as the fitting condition to speed up the extraction speed of the
boundary line. Experiments show that the average recognition accuracy of the method reaches 95%, the average time of
navigation path extraction is 30.5ms, the average image processing time is 0.89s/frame, and the extraction results are
used for visual navigation, the relative error of visual navigation is less than or equal to 12.8cm at the operating speed of
0.8m/s, and the relative accuracy of navigation reaches 95.07%, the algorithm under the complex environment of paddy
field has good applicability and high real-time performance, and provides effective navigation information for
autonomous navigation of combine harvester.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Criticality is a measure of the difficulty of video compression. Image sequences with high criticality are required for the
evaluation of video compression algorithms. The selected test sequences are usually determined by experts from large
amounts of material,which is labor-intensive and subjective. In order to solve this problem, a test sequence selection
algorithm for video criticality evaluation is proposed in this paper. Based on basic principles of video coding, four types of
metrics including texture map variance, AC energy, motion vector difference and motion angle entropy are selected in this
paper. The least squares method is applied to fit the parameters of these four types of metrics to their corresponding
criticality values for multiple resolution image sequences, and the fitting results are verified by experiments to effectively
select the test sequences.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
In this paper, an adaptive aggregation stereo matching algorithm based on edge preservation is proposed. The proposed
algorithm uses the edge space information of the image to construct a weight matrix, highlights the role of the straight
line in the edge weight matrix. Furthermore, the algorithm introduces the idea of cross aggregation, and utilizes the edge
weight value to improve the window arm length determination function. Experiments are carried out on Middlebury
stereo matching platform, and the results show that the proposed algorithm can distinguish the cost of pixels in different
regions more effectively, and finally can improve the matching accuracy of the images in all regions.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Under the background of digital informationization, the visual culture transmitted by digital information image is
constantly changing and developing, which also brings the change of visual experience and aesthetic concept. Because of
the complexity of plate making and the limitation of printing, the dissemination of traditional black and white prints has
been seriously affected. For this reason, the technology is on the premise of fully mastering the characteristics of black
and white printing, combined it with computer technology, relying on Photoshop software to put forward a set of
ordinary image materials into black and white digital prints, through plate making to solve the problem, print, save and
spread, then verified the effectiveness of the program through experiments.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
The feeding volume parameter of combine harvester greatly affects harvesting efficiency. Therefore, this paper proposes
a feeding volume prediction method. This method uses the camera installed at a certain height above the front of combine
harvester to capture the images of crops during operation. Then, the ARM processor unit does image processing and BP
neural network inference. Here, the image processing includes color space conversion, histogram equalization, filtering
operation and other operations. And its output is the pixel value of rice panicle layer. As for the neural network, the pixel
value, rice stubble height, moisture content and grass to grain ratio are used as the input parameters, and the output of
network is feeding volume. The result of neural network shows the decisive factor of predictive model up to 0.95, which
is able to predict the feeding volume properly. At last, according to the result of experiments, this method predicts feeding
volume well, and the relative error of the predicted average feeding amount is 10%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Three-dimensional point cloud registration is a research hotspot in the field of computer vision. The shape of the point
cloud is an essential form of describing the point cloud. Many registration algorithms form feature descriptors that can
describe the shape by extracting features. These feature descriptors express the difference in the shape of the point cloud.
This paper proposes a new feature descriptor based on quadratic error, which has rotation invariance. Two feature
descriptors are obtained by extracting the quadratic error of the point and the quadratic error of the neighborhood point.
Although the point cloud registration algorithm based on Gaussian mixture model has good robustness in terms of noise
and outliers, this type of method does not perform well for larger-scale rotations. In this paper, the quadratic error feature
descriptor is used as the local features of the Gaussian mixture model to optimize the registration effect of the Gaussian
mixture model in larger rotations, and a dual-feature-based registration strategy is proposed to optimize the defects of
single-feature-based registration. The experiment compares our proposed algorithm with the robust ICP algorithm and the
popular feature-based registration algorithm regarding registration efficiency and registration accuracy. The efficiency is
3-4 times that of robust ICP. In large-scale rotation, our proposed algorithm has good robustness and is superior to
mainstream algorithms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
The classification and detection of glomeruli of kidney tissue is a key process for correct diagnosis of diseases in renal pathology. However, it is still a big problem to perform comprehensive and accurate glomerular ultramicroscopic pathological diagnosis based on high-resolution whole-field digital slices (whole Slide Image). The reason lies in the grayscale image texture corresponding to the glomerular ultrastructure. Complicated, there are many types of related lesions, and it is difficult to identify and judge subtle pathological changes. The traditional semantic segmentation model cannot achieve the ideal segmentation effect. Based on this, this paper proposes a semantic segmentation method FEU-Net (FCN-Efficient U-Net) for pathological slices, which improves the accuracy of glomerular region segmentation and realizes end-to-end segmentation. FEU-Net uses EfficientNet, which has undergone transfer learning, as the encoder part to enhance image feature extraction capabilities. The decoder uses U-Net to promote the fusion of deep and shallow features while reducing the amount of network parameters, and redesign the convolution module to improve the gradient transfer capability. Compared with other classic methods in the SEED data set, it verifies that some classic models are difficult to fit in this segmentation task. At the same time, experiments show that modifying the feature extraction method can greatly improve the results. Among them, the modified method in this article is in the accuracy of segmentation. This is an increase of 18.968.96% over the original U-Net. At the same time, this article conducted ablation experiments on the ZENODO data set and the mendeley data set, verifying that each module in the improved algorithm helps to improve the segmentation effect of pathological slices. In the SEED data set, the FEU-Net method in this paper improves the Dice coefficient, accuracy rate and recall rate by 5.175.17%, 2.72.7%, 3.693.69%, 4.084.08% respectively compared with the benchmark model; in the BOT data set, The three indicators of the method in this paper have increased by 0.47%, 0.060.06%, 4.304.30%, and 6.086.08% respectively. The FEU-Net proposed in this paper improves the accuracy of segmentation of the pathological slices of gastric cancer and has good generalization performance.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
While the new pneumonia epidemic (CVD-19) has had an unprecedented impact on the traditional economic model, it
has also given a huge boost to the new retail industry immersive technologies such as Virtual Reality (VR) and
Augmented Reality (AR). Mixed Reality (MR)will achieve breakthrough development of even more 5G technology,
focusing on digital media with multi-sensory interaction features such as visual, auditory, and kinesthetic-immersive
media-to build a new paradigm for commercial information dissemination.Therefore, it is of obvious relevance to
understand the value of immersive media in the digital retail space an to find the right way to apply it to the way coastal
cities grow economically.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Visibility prediction is a concern issue in the field of public transportation, which is related to the normal operation of flights and the safe travel of vehicles. Therefore, reasonable prediction of visibility is very important to improve the efficiency and safety of public transportation. This paper mainly combines data and video, and images are quantified for quantitative analysis of large fog evolution trends. Further, the mathematical model is established to study the visibility prediction problem that is concerned with the current traffic, aviation field, and proposes targeted recommendations for the current predictive means. Preprocessing the sample data, eliminates the wild value, performs interpolation processing on the missing position, regression analysis of the image, obtains the regression model of image visibility changes, performs visibility prediction, select accuracy, adaptive ability, good depth integration Convolutional Neural Network (CNN) is an algorithm for learning processing on the image. In the model establishment, the three image processing modes (Fourier change algorithm, spectral filtering, original pictures) are established, and the hidden feature, depth integrated volume is established by the three image processing modes (Fourier change algorithm, spectral filtering, original pictures). Total Neural Network (CNN) simultaneously learns three images of the input and generates a classification output, and finally the categorized visual network model.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Domain adaptation has aroused heated interest in medical image analysis due to the universality of cross-modality data, e.g., CT and MRI. Among that, unsupervised domain adaptation has become increasingly important because of the lack of high-quality manual annotations. Deep learning methods have demonstrated the state-of-the-art performance on the above tasks, especially the adversarial learning methods such as Synergistic Image and Feature Alignment (SIFA) network. Based on the elegant benchmark SIFA, this paper presents an improved unsupervised domain adaptation method by introducing a multi-task branch for target image reconstruction. The network is implicitly improved to learn domain-invariant features via the image-level alignment in image reconstruction space. We achieve 82.8 Dice and 4.7 ASD on the 2017 Multi-Modality Whole Heart Segmentation Challenge dataset, demonstrating that our method is effective in improving segmentation performance on unlabeled target images.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
The statistical characteristics of industrial big data contain the dynamic failure information of equipment, which is
related to the operation safety and maintenance performance of equipment, especially the randomness of recessive or
intermittent faults, the instability of state and the uncertainty of duration, which are the key factors leading to the
changeable abnormal state and sudden failure of equipment, and are difficult to detect and diagnose. Therefore, aiming at
the industrial data source fusion technology, this paper analyzes the information allosteric effect of unit structure visual
image, and explores the physical meaning and identification method of characteristic parameter representation of
equipment unit structure visual strain image; it is helpful to improve the recognition accuracy of hidden or intermittent
fault features of equipment by optimizing the algorithm of fault feature recognition based on edge computing and
clarifying the key theoretical problems of dynamic fault feature recognition and prediction algorithm based on edge
computing. The above research has laid a theoretical foundation and technical support for improving equipment
reliability level and ensuring safe operation and maintenance efficiency, and has important social benefits and
engineering practical value.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Coherent Point Drift (CPD) is one of the popular robust point cloud registration algorithms in recent years. However, the
algorithm uses fast Gaussian transformation to calculate the matrix-vector product, resulting in slower overall registration
efficiency. We propose an improved coherent point drift (ICPD) algorithm, which introduces faster Gaussian lattice
filtering to calculate the above product and uses the global squared iterative method to reduce the number of iterations of
the CPD algorithm. In addition, the outlier w is not accurately expressed in CPD. We propose an iterative outlier formula
to solve this problem. Experiments show that the improved algorithm is about two orders of magnitude faster than the CPD
algorithm, 1-2 times faster than the ICP algorithm, and shows superior performance in environments with different noise
and outlier distributions.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Most of the existing camouflage effect assessment methods focus on analyzed the characteristic differences between
target and background, which are relatively single. This paper takes the camouflage target in the video as the research
object, proposes a moving target camouflage effect evaluation method based on the detection of the camouflage moving
target and the detection time constraint. The method mainly includes two modules: camouflage target detection and
camouflage effect evaluation. It uses image layering and feature matching to obtain high-precision detection results,
combines the running time of the target matching algorithm and the time required for the observer to locate the target to
construct the camouflage effect evaluation index system, and comprehensively evaluates the camouflage effect of motion
target from the perspective of detection accuracy and exposure time.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
In order to solve the problem of selection of the shipbuilding demander for the supplier in cloud manufacturing environment, an evaluation and selection method based on comprehensive demand value is proposed from the perspective of the supplier. Firstly, the evaluation index system of the demand side is established, and the demand capability is obtained by using the variable precision rough set model. Secondly, considering the supply preference of the supplier, the preference coincidence degree of the demander is obtained combining with the demand information of the demander. Finally, the comprehensive demand value is obtained by integrating the two factors, which can provide reference for the selection of the supplier.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Remote Sensing Image Recognition and Optical Imaging
As the difficulty of measuring infrared radiation characteristics is increasing, the requirements for target recognition
technology are increasing. Image fusion is a key issue in target recognition. In this paper, an image fusion algorithm
based on the ORB features point matching is proposed, and the infrared image is experimented. The experimental results
show that the proposed image fusion algorithm has good performance in the fusion efficiency and fusion precision, and
can be applied to the identification of infrared target.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
In recent years, with the continuous development of remote sensing technology and computer vision technology, the semantic segmentation of remote sensing images is of great significance in terms of earth observation, urban planning, military simulation, etc. This paper proposes a remote sensing image semantic segmentation method based on improved Deeplabv3+. Firstly, the backbone network is improved. Xception is selected to replace the traditional ResNe101 as the backbone network for the improved Deeplabv3+, and the network structure is deepened and depth separable. Optimization methods such as product replacement improve the segmentation efficiency; then, in order to improve the feature extraction effect of small targets in remote sensing images, the expansion rate of the cavity convolution in the ASSP module is optimized and adjusted. The experimental results show that the improved Deeplabv3+ algorithm has achieved good segmentation results on the data set, miou reached 91.23%, pixel accuracy reached 93.31%, and F1-score reached 89.2%, which is an increase of 2.4%,1.9% and 2.7% compared with the original Deeplabv3+. At the same time, compared with mainstream U-net and SegNet algorithms, this algorithm also has strong advantages in semantic segmentation of remote sensing images.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
This essay discusses the important role of deep learning technology in optimizing the mobile visual search function of
digital library. On this basis, it analyzes the necessity and feasibility of constructing mobile visual search mechanism based
on deep learning technology, and expounds the application of deep neural network to construct mobile visual search. The
main content of the visual search mechanism, including mobile visual big data input layer, mobile visual resource
organization layer, deep neural network analysis layer, and mobile visual service interaction layer, etc., are designed to
enhance the value of mobile visual search in a big data environment and promote big data The realization of personalized
intelligent services of the Times Digital Library.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
In order to improve the rapidity and real-time of infrared image generation, and make the user and computer have a good interaction, the method of developing virtual reality simulation system based on Vega Prime environment is introduced. The 3D model of target is established by Creator software, and the infrared and atmosphere are modeled by TMM and mat tools of Vega Prime. At the same time, the attenuation effect of atmosphere on infrared radiation is considered. The infrared image is simulated with Ondulus IR software module of Vega Prime. Ondulus IR is a new simulation module, which is based on the accurate infrared radiation principle for modeling and real-time simulation, and real-time rendering to generate infrared images. When using Creator to model, the principle is to reduce the number of polygons as much as possible. The completed model needs to be simplified by deleting polygons and merging faces, which improves the realtime and speed while maintaining the characteristics of the model. Based on VC + + 6.0 integrated development environment, combined with the development characteristics of MFC application program, this paper analyzes the development environment of Vega Prime software, realizes the simulation process and outputs the results of infrared image simulation, which has certain practicability in the development and implementation of infrared visual simulation.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Aiming at the problem of spraying target identification and positioning of steel structure fireproof coating spraying robot,
the stereo imaging model of binocular camera is established, and the collected binocular images are preprocessed by
distortion correction and epipolar rectification. The application of multi-grid method in stereo matching and depth image
calculation were studied, and the effect of multi-layer perceptron neural networks classifier (MLP) on segmentation of
steel structure region in color image was researched. The experimental results in real scenes show that the error between
the binocular vision measurement results and the actual distance is less than 2.5%. MLP classifier can effectively
separate the target region from the color image, and further crop the depth image in this region to obtain the depth image
containing only the target information.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Personal identification is a very important technique in the computer vision. Recognition of biological feature (e.g., lip)
is broadly applied because of its uniqueness and convenience. This paper proposed a framework for lip print recognition
using Local Binary Pattern based on bit plane (BLBP). The BLBP features were first obtained from lip print. In order to
achieve robust recognition, the BLBP features were normalized and Support Vector Machine (SVM) with Gaussian
radial basis function was used for classification. The results showed that the proposed framework was more effective and
accurately comparing with other methodologies.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Thermal interference will have a certain impact on the response performance of flame detectors. In order to evaluate the
anti-thermal interference performance of the detector, it is necessary to study standardized thermal interference
simulation test methods. Common thermal interference mainly stems from two types of conditions: the heating of
various electrical equipment, pipes and containers in industrial sites; the heating of the detector itself or objects within its
monitoring range after being exposed to sunlight. Based on the analysis of the both types’ thermal interference
characteristics, and using Planck formula and the Stefan-Boltzmann law, the thermal interference equivalent blackbody
radiation model and test methods of different severity levels are constructed. Experimental tests verify the feasibility and
effectiveness of the method.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
With the continuous advancement of computer technology, Chinese style digital illustration has shown more and more appeal in different fields such as packaging and posters. Adobe Photoshop is a powerful digital software, and a considerable part of Chinese style digital illustration is drawn by it. The texture of the digital illustration plays an important role in showing the characteristics of Chinese style. With its powerful function settings, PS brushes have become effective means to create this kind of texture. Through the analysis of the texture of Chinese style digital illustration, the article points out the main influencing factors in the creation of unique texture, discusses the advantages of PS brushes in expressing Chinese style, and summarizes the key points of using. From the perspective of texture shaping, the digital illustration can better promote the Chinese style.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Line detection in images has always been one of the important research fields of artificial intelligence computer vision
for image processing.Hough transform (HT) is one of the most extensive detection algorithms in image processing and
machine vision applications, and has the advantages of anti-noise. However, it involves a huge amount of computation
and too much memory.In view of its shortcomings, it is recommended to improve its performance. In this paper, a new
detection method is cited, and the statistical probability Hough line transformation is added.The first step is to use the
Canny operator to obtain the edge information of the gray image and use the improved method to detect the straight line
segment in the image. This algorithm enhances the accuracy of edge detection and optimizes the extraction of the
straight line. At the same time, it has the simplicity and wide applicability of Hough transform.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
In this article, we innovatively use Pearson correlation coefficient, etc. to analyze the components of the snowy image (the
snow-free image and the mask image), and then create an effective and accurate snow-free model using the relationship
between the components of the snowy image. For the snow-free model, we innovatively consider the similarity in the
generation of the snow-free image and the mask image; we also consider this relationship in our neural network framework.
We set the generator model in the generative adversarial network as the pseudo-siamese network with the same structure
but the different parameters. Each branch of the pseudo-siamese network adopts the autoencoder and the multi-scale
perception structure. The former can guarantee the acquisition of high-resolution images, and the latter can perceive context
information at different scales. We restore the mask image first and then restore the snow-free image because the mask
image has a simpler background than the snow-free image and the mask image is easier to recover than the snow-free
image. The results show that our (PS-GAN) pseudo-siamese generative adversarial network not only has better
performance on the data set, but also has good results in the real world which greatly improves the effect of using the
yolov5 to detect.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
During the use of a convolutional neural network to train a recognition model of plant leaves, the convolutional layers
focus on the appearance of leaves in learning the features of them, while ignoring their internal texture features, thereby
resulting in the misclassification of plant leaves with similar appearance. Aiming at this problem, this paper proposes an
accurate identification method of plant leaves based on multi-feature fusion, which can be applied to extract the
appearance and texture features of leaves simultaneously, and to conduct fusion and summation for these two types of
features. The experimental results indicate that compared with the accuracy of the ordinary convolutional neural network
recognition method and traditional machine learning method, the accuracy of this method has been improved
substantially.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Hyperspectral image(HSI) classification is a research hotspot for its wide application. However, to obtain labeled HSI
image data is time consuming and with high cost, which makes it difficult to design an optimal classifier. In this paper, we
propose a novel HSI classification method based on active learning. Before the iteration begins, training patch set is used
to train a CNN model. Based on the classification result of current model, a carefully-designed patch selection strategy is
employed to select patches. In each iteration, we fine-tune the current CNN model with the selected HSI patches. Repeat
the iteration until the final classification accuracy is satisfactory. Comprehensive experiments on the University of Pavia,
Indian Pines, and Salinas demonstrate the effectiveness of the proposed method.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
This paper proposes a face attribute transfer system based on a deep convolutional network and a generative adversarial network, Self-recognition Style-encoding Face Editing Network(SSFEN). The network only needs a face source image to edit the facial features of the image. The whole network consists of two modules: self-recognition style-encoding network and multi-styles transfer network.The self-recognition style-encoding network aims to learn the style features of all faces in the dataset according to the input initial face image and features, and through further analysis and processing of the deep convolutional network, it outputs the facial feature coding of the original image. The multi-styles transfer network draws on the idea of generating a adversarial network, and only trains one generator to complete the style transfer in multiple fields, and adopts the generator to complete the editing of each field of a single face source image. In the joint training of the self-recognition style-encoding network and the multi-styles transfer network, we mainly apply the multi-label and multi-class discrimination loss, the adversarial attribute loss, the style domain recognition loss and the cyclic loss of the target domain. In addition, the style domain isolation loss is proposed to reduce the mutual influence between various target domains when the face is edited in a single target domain, which increases the accuracy of facial feature editing.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
As for the problem of low SAR image target recognition rate, the present paper improved the single module of residual
attention network. Firstly, max pooling, average pooling and stochastic pooling were combined to put forward a dynamic
hybrid pooling method for inaccurate middle and down sampling of mask branch to make the weight of hybrid attention
of mask branch extraction more accurate; and then, channel attention mechanism was added to the trunk branch to
enhance the weight of the useful feature, so as to improve the efficiency of information flow. The experiment based on
MSTAR data set indicated that, compared with other algorithms, the improved model was relatively accurate. The
accuracy was 1.16% higher, at the same time, the size of the improved model was only 1/3 of residual attention network.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
The composition of the soil is very important, therefore, it is necessary to separate soil and other components from soil
images in order to facilitate the study of soil components. This paper mainly studies the realization method of soil image
segmentation, especially the traditional maximal inter-class variance method in global threshold method. On this basis,
the fuzzy C-means clustering algorithm is combined with fuzzy theory to optimize the algorithm. By comparing the
experimental results, it is proved that the fuzzy C-means clustering algorithm based on the maximum inter-class
segmentation method can achieve the segmentation of objects and backgrounds, and meet the requirements of image
segmentation.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
At present, with the improvement of the quality of life, people's requirements for the quality of food are also gradually
improved. The most effective way to effectively supervise the quality of food is to carry out food traceability supervision.
Fish can provide people with abundant protein and excellent taste, and it has gradually been favored by people. However,
due to the complex underwater environment, RFID technology, which is often used for pigs, cattle and other animals, is
not suitable for large-scale fishery breeding, and it cannot be effectively traced to the source of fish food, which is difficult
for the government and consumers to supervise. Using deep learning method to identify animals can avoid a large amount
of labor costs and effectively trace the source of fish. Taking golden crucian carp as an example, this paper proposed to
use YOLOV5 to detect the position of golden crucian carp and cut it out, and applied Siamese Network to carry out
identification, and the identification accuracy reached 89.8%, which provided research ideas for the identification
technology of fish.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
With the rapid development of wireless communication technology, modern electronic map has gradually changed to many
aspects, such as information transmission, map display, dynamic map and so on. Therefore, the traditional two-dimensional
map can’t meet the needs of society, which needs to constantly improve the online dynamic map. Based on browser
technology, WebGL technology has become a new tool, which can help the web to realize new map engine functions.
Based on GIS technology, we can design a real-time, efficient, scalable, safe and reliable map engine. Through WebGL
technology, we can draw massive vector data, which can also improve the loading speed of the server. Firstly, this paper
analyzes the function of online dynamic map. Then, this paper lists some algorithms. Finally, this paper constructs an
online dynamic map service framework.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Computer vision-based methods have been extensively used in vibration measurement of engineering structures in recent
years. However, vibration testing of engineering structures usually requires both the high accuracy for low displacement
and robustness of the modal test, which is difficult for most vision-based methods. In the current work, we propose a
vibration displacement extraction method based on point tracking, Tracking-Detection-Deformation Matching (TDDM),
which combines the advantages of both the detection and tracking. Firstly, TDDM used the tracking method to obtain the
initial local motion region; Then, the tracking error in the local region was corrected by the detection method; Finally,
based on the correction result, the local deformation matching was designed to improve the signal-to-noise ratio of low
displacement. The effectiveness of the TDDM tracking algorithm is verified by vibration experiments of Bench-scale
building structures. The experimental results demonstrate the robustness and high accuracy of the proposed method in
structural vibration measurement.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Image visual try-on aims to transfer the target clothing image to the reference person, and it has become a hot topic in
recent years. Current techniques usually focus on preserving the character of clothing images when warping it to
arbitrary human pose. However, it is still a challenge to produce realistic try-on images when the reference person has a
complicated pose. We propose a novel repair model to solve this problem. We train the network by simulating arm
breaks and improper occlusion, so that it can be automatically repaired. Compared to the newest advanced techniques,
we improve upon the inaccuracy in the predicted semantic segmentation, proposed a new generator model to obtain more
detail from images.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Salient object detection is still a hot topic at the moment. Based on the pixel-by-pixel prediction of the image, the salient
object detection model often has a large number of parameters, which brings problems such as high latency, large-scale
computing and storage difficulties. In this article, we try to explore the application of the idea of structural reparameterization
in real-time salient object detection, trying to alleviate the contradiction between the above three
problems and accuracy and focusing on the important feature of optimizing the reasoning speed of the network. Based on
the above ideas, we propose a plug-and-play flexible convolution with a new structure. The convolution module can be
converted to a normal 3x3 convolution layer in the inference stage, so the inference speed will be greatly improved.
Using the plug-and-play convolution module and pooling together, we construct a model with high inference speed.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Because the small target image has fewer pixels, it is prone to miss detection and error detection; the target detection
algorithm needs to be improved and enhanced to deal with the problem about the small target detection in specific
scenarios. In this paper, small target detection algorithm is applied to the new field of illegally modified vehicle detection
to reduce the workload of traffic management. The commonest modification is to install a rear wing and change the contact
angle between the hub and the ground. This paper proposes a method to detect modified car parts based on improving
Faster-Rcnn, in order to detect two modifications described above. On the basis of the original Faster-Rcnn, using
multi-scale training and increasing the number of Anchors to enhance the robustness of the network in detecting targets of
different sizes, and introducing the Soft-NMS algorithm to replace the NMS algorithm, to solve the problem of partial
overlap between two targets when the distance is close, the possibility of missed detection of targets with low confidence
and bounding boxes with high confidence scores are not always more reliable than bounding boxes with low confidence.
Experiments show that compared with the original Faster-Rcnn, the detection accuracy is increased by 4.6%, and the
model has a certain generalization ability and robustness.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
At present, integrated architecture is adopted by orchard fertigation management systems. The coupling of the system is too strong, which makes its operation and maintenance cost too high. To solve the problem, this article designs the architecture of the software platform under the cloud computing model. The resources related to the management of orchard fertigation are built into cloud services. These cloud services are described in categories and scheduled uniformly. These services are combined with the support of expert strategies. The corresponding business process(BP) is established to realize the intelligent operation of orchard fertigation.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
In the continuous prevention and control of the COVID-19 epidemic, community work has become an important basic
unit for epidemic prevention and control. The prevention and control of the epidemic is inseparable from the
maintenance of the community's health environment. Whether garbage can be properly and timely handled is the key to
epidemic prevention and control. This research is based on the national and Tianjin municipal garbage classification
policies, and analyzes the problems and difficulties of community waste classification during the epidemic period, and
explores the auxiliary interactive design plan of garbage classification suitable for the "Jingxianli" community under the
background of community epidemic prevention and control.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Through the thinking and exploration of digital sculpture based on sculpture creation, analyze the element composition
and new forms of digital sculpture, explore the new language characteristics of digital sculpture, study the application of
digital sculpture, and look forward to the digital sculpture based on the background of new technology. Making some
meaningful attempts in the creation, processing, and display process of the new technology is helpful for the entry point
and fit of the digital sculpture software in the new technology, and for the objective understanding of digital sculpture.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
This paper proposes a method based on channel pruning and generalized intersection ratio. This method focuses on
solving the problems that the deployment of the large volume model is difficult and the original loss function cannot
reflect the real overlap between the real frame and the predicted frame. The proposed target detection method was
compared with the YOLOV3 network model, the R-CNN method and the Faster-CNN method in the VOC2007 dataset.
The results show that the proposed target detection method has smaller model size, faster inference speed, higher
accuracy and less dependence on computer hardware, which is more suitable for deployment in special environment such
as underground coal mine.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Explore the application strategy and development direction of virtual reality technology in skydiving experience design.
Through the analysis of virtual reality technology, the basic concept, basic characteristics and basic composition are
explored, the system design scheme based on virtual reality technology and LOD technology is proposed, and the overall
structure of each part of the system, the composition and construction of hardware equipment, the function of each
module of the software system and the important technology are demonstrated and analyzed. On the basis of this
research, the feasibility of combining skydiving experience design with virtual reality technology is summarized. It is
concluded that this design can break the limitation of space and time in skydiving experience, and avoid the danger
brought by operation in real environment, so as to improve accuracy and economic benefits. Virtual reality technology is
the product of the rapid development of computer technology, and the combination of skydiving experience design has
realized the combination of science and art.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
It is very important to improve the purity of classified waste from the source for the recycling of waste resources and the protection of ecological environment. However, the current garbage sorting sites often require manual visual inspection, which is labor-intensive and unreliable. Recent achievements in convolutional neural network make deep learning-based visual detection technology provide a new way for intelligent development of cities. In this paper, a systematic design of the intelligent garbage detecting system and corresponding workflow are implemented to reduce labor costs. An improved Cascade RCNN algorithm, which can greatly improve the detection accuracy, is proposed for garbage detection of this intelligent device. Furthermore, in order to solve the problem that some types of garbage data sets are less, we use generative adversarial network to implement data augmentation. The experimental results show that compared with original Cascade RCNN and other commonly used target detection networks, our proposed method performs better at garbage detection.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
Compared with traditional sculpture design and production methods, digital technology has been widely used in our
country, such as games,film and television, animation. As a digital software, ZBrush has been widely used by sculptors.
In the creation process, some artists use it as a tool to create new sculpture art, try art creation methods that have not
been explored before, explore the language characteristics of digital modeling different from traditional creation methods,
and broaden the art creation techniques and creative thinking. Lucid traditional carvingPlastic creative ideas have a
promoting effect[1]. Especially in the process of modeling and creation using ZBrush, you can explore the unique artistic
language of digital sculpture, and play the charm of digital materials in multiple directions. Art creators can use ZBrush
software to realize the reproduction and simulation of the material, texture and space of sculptures in the virtual space, so
that the creators can create freely in the three-dimensional virtual space and exert unlimited possibilities.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
The current stereo matching network models have excellent performance in the aspect of matching accuracy, but most
models have large number of parameters, which is difficult to meet the requirements of real-time performance, and
reducing the model size will lead to the decline of their accuracy. In order to improve the accuracy and real-time
performance of stereo matching, a real-time stereo matching model based on deformable convolution was proposed. In
order to improve its accuracy, the model adopts 2-dimensional deformable convolution to extract more effective features,
and adopts 3-dimensional deformable convolution to process the matching cost volume after processing, aggregates the
image pairs from the disparity and spatial dimension more effective correlation, forms a better matching cost volume,
and uses spatial propagation network to optimize the disparity. In order to improve its real-time performance, the model
adopts the method of residual learning to reduce the amount of memory and computation consumed in forming matching
cost volume. The prediction errors are divided into three stages. The endpoint errors of the three stages on the SceneFlow
dataset are 2.99, 2.26 and 1.99, respectively.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
In the process of sewage treatment and dosing, due to the physical and chemical factors of the medicament or thinking
factors, the market will cause the dosing abnormal. To monitor the abnormal situation in real-time, this paper uses the
Long Short-Term Memory algorithm to learn the time law of dosing and uses real dosing data to predict the value of the
next time point by point, which can determine whether an exception has occurred in the dosing system reliably.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
The paper optimizes the YOLOv4 detection model incorporating on the unmanned vehicle detection scenario. A hybrid
attention mechanism combined channel and spatial attention mechanisms is introduced into the backbone network of the
detction model. The high level semantic feature map and the shallow fine-grained feature map are merged into the
detection neck network. The hybrid attention mechanism strengthens the process of screening the fine-grained features.
The effect of using hybrid attention mechanism is better than that of using channel attention mechanism. Through the
experiments, the results were drawn: (1) For the CSPDarknet-53 network, the effect of using hybrid attention mechanism
is better than that of using channel attention mechanism. (2) On VOC 2007+2012 dataset, the model combined shallow
layer feature maps and hybrid attention mechanism in this paper can improve can improve the detection accuracy by
about 4%, compared with the original YOLOV4 object detection model. At the same time, for the KITTI dataset images
closed to the actual scene, the image are used to verify the actual driving scene. The improved model achieves excellent
detection effect on KITTI dataset.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
The zinc sulphide concentrate fluidized bed furnace is used to Oxidize sulphide in zinc concentrate to produce the zinc calcine. The structure of the wind box in this furnace is plays the crucial role to decide the fluid flow in the furnace. The numerical model of the single-phase flow in the furnace was established using the standard k-ε model. Two different sets of the wind box structure were built. The single-phase flow of the velocity field in the wind box of the zinc roast fluidized bed furnace shows that the perforated plate can evenly distributed the air velocity to get better fluidized conditions. The detail flow behaviour was showed, and it will help the designer to understand the velocity field inside the furnace.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
In order to better realize the effective regulation of building refrigeration, the optimization design of energy regulation
model of multiple refrigeration units is carried out in combination with cold and heat inertia, and the data acquisition,
regulation and management of partial load parameters are carried out in combination with refrigeration unit frequency
conversion technology, so as to ensure the rationality of refrigeration unit energy regulation and the design goal of
building energy conservation and emission reduction. Finally, it is confirmed by experiments. The capacity regulation
model of multiple refrigeration units with building cooling and heating inertia proposed in this paper has high rationality
in practical application and fully meets the research requirements.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
In order to ensure the stable and safe production and operation of dispatching data network, a dynamic monitoring
method of regional communication network service quality for digital city construction is proposed. The network quality
monitoring technology is applied to dispatching data network. Probe equipment for network quality monitoring is
deployed at the exit side of each service area of backbone network and access network to implement end-to-end
real-time detection of services. The low efficiency of manual testing, troubleshooting into efficient, intelligent real-time
monitoring and positioning. Through simple reports and charts, we can directly evaluate the network operation and
improve the efficiency of operation and maintenance.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.