Industry has traditionally relied upon point sensors such as thermocouples and pressure gauges to control its operations, but as manufacturing processes become increasingly complex, additional types of information are required. For example, typical process measurement needs now include contamination detection, particulate size and shape, concentration/density profile (in pipes and tanks), and thermal profile. In many cases, imaging systems are the only sensors that can provide the required information.

Methods for optical access to combustion chambers of internal combustion engines are shown. The optical access is either maximized to allow application of complex optical diagnostic techniques while maintaining minimum necessary operability of the engine or engine components, or full engine operability is maintained while optical access and diagnostic techniques are tailored to the diagnostic demand and the restraints of engine operation. As full optical access allows a wide range of diagnostics to be applied, it puts the emphasis on organizing and extending realistic engine operation conditions. Full engine operability, however, challenges the optical access and the design of diagnostic systems. The paper gives examples for diagnostic applications for both of these situations.

The spatial distribution of chemical species can be a critical determinant of chemical reactor performance. The spatial variation of air-fuel ratio in a combustion chamber of an internal combustion engine has significant influence on fuel efficiency and emissions. We report the development of a fiber-based near-infrared absorption tomography system, to measure the distribution of hydrocarbons in cylinder. It has been successfully applied to transient gas injections. The technique exploits the specific (but weak) hydrocarbon absorption of 1.7 mm radiation, which wavelength has only recently become accessible by availability of solid-state alloptoelectronic components. A standard telecommunications laser was also deployed to measure reference information. The measurement space is sampled by 32 dual- wavelength fiber-coupled measurement paths. The logarithm of the ratio of the two measurements yields the path integral of the hydrocarbon absorption, and hence, of concentration. The path integral is measured with typically 28 dB signal-to-noise ratio. Single-channel characterization shows that the technique is readily calibrated for temperature and pressure effects, over the region 70– 150 °C and 1–10 bar. Tomographic reconstruction of different gaseous hydrocarbon flows has been achieved with spatial resolution of the order D/5, where D is the vessel diameter. Temporal resolution of about 1000 frames/s is demonstrated.

Magnetic resonance imaging (MRI) is a noninvasive method for quantitating in three dimensions the spatial distribution of water and/or hydrocarbons in an optically opaque environment. This article uses three applications to illustrate how that unique ability of MRI can provide insight to aspects of process engineering. The first involves mapping temperature distribution due to microwave and/or conventional heating. The second demonstrates how MRI can be used to measure the flow of fluids through complex geometries, including a screw thread extruder and a porous medium. The third, measurement of solid–liquid separation, is illustrated by studies of two different types of filters. Although no background to the theory is provided, substantial practical details are given about the design of MRI-compatible processing equipment.

The paper surveys the status of electrical tomography for industrial applications. For the present purposes this is considered to include ‘‘low’’ excitation frequencies, up to about 1 MHz, comprising resistance (electrical resistance tomography), capacitance (electrical capacitance tomography), inductance (electromagnetic tomography) and impedance (electrical impedance tomography) modalities. Introductory background material is followed by comparisons between the instruments that have emerged. Data processing is considered with emphasis on the trade-offs that are necessary in implementing algorithms for reconstructing images. Recent applications are tabulated and case studies are presented for six contrasting areas that illustrate significant progress towards industrial benefit. Some comparison is made with applications to medical tomography and a number of issues are identified for future research.

It is highly desirable to be able to interrogate in three dimensions the whole fluid inside a typical stirred vessel, especially when used as a chemical reactor. Electrical resistance tomography can scan the fluid with a spatial discrimination of the order of O(103) voxels with video frame rate acquisition speeds. Some new augmented-reality results are presented for a 2.3 m3 pilot scale vessel detecting miscible fluid mixing of a brine tracer pulse and solid– liquid mixing. The images are created by solid-body iso surfaces between which are variable opacity layers. Extra augmentation can be achieved by the use of color, giving a five-dimensional (5D) representation of the mixing process. For miscible mixing, the geometry and spatial patterns clearly highlight the effect of injection position on mixing rates. For the solid–fluid mixing, it is possible to visualize in 5D the way solid resides on the vessel base behind baffles as well as indicate the height to which solids are effectively suspended. The results have also been represented as conventional axial solids distribution profiles. The dynamics of mixing after a charge of solids can also be captured and visualized using augmented reality. The facility to envisage mixing information in up to 5D offers new possibilities for control action linked to spatial distributions of phases and components.

In this paper we consider the reconstruction of rapidly varying objects in process tomography. The evolution of the physical parameters is approximated with stochastic convection diffusion and fluid dynamics models. The actual time-varying reconstruction is carried out as a state estimation problem. As the boundary observations we use the voltage data of electrical impedance tomography. We have previously shown that state estimation works well in process tomography in the cases in which the fluid dynamics of the system are modeled correctly. In the real case, however, the velocity field cannot usually be determined accurately. This may be caused, for example, by complex nature of the flow, the turbulence, discretization, etc. In adopting the first proposed approach, it is essential to know how much the inaccuracies in the fluid dynamical model affect the state estimates in process tomography. In this paper we consider the tolerance of the approach with respect to these inaccuracies. We show that the estimation scheme is relatively tolerant to modeling errors in the flow field. Thus relatively reliable estimates can be obtained, for example, in a case in which a laminar flow model is used in turbulent flow conditions. However, the degradation that is due to incorrect flow fields is not insignificant and it is also conjectured that it could be possible that an extension of the proposed method could be used to estimate some flow field parameters.

Due to practical considerations, electrical fields and other soft field sensing techniques are convenient to use in process tomography applications. However, most of the reconstruction algorithms were first developed for hard sensing fields that approximate soft fields only in the situation of low electrical contrast. A less restrictive reconstruction method can be developed by refining the qualitative images of a direct imaging probe, but the boundary measurements must be sensitive to changes in the distribution of the electrical properties of the flow. This paper addresses the issue and presents a sensitivity analysis of different excitation strategies and their applicability in a reconstruction algorithm such as described above. Results confirm that classical excitation strategies suffer from a major lack of sensitivity and that new ones must be developed, possibly based on the optimization of the excitation profiles or on multisensing techniques.

The paper deals with the goal of component fraction estimation in multicomponent flows, a critical measurement in many processes. Electrical capacitance tomography (ECT) is a wellresearched sensing technique for this task, due to its low-cost, nonintrusion, and fast response. However, typical systems, which include practicable real-time reconstruction algorithms, give inaccurate results, and existing approaches to direct component fraction measurement are flow-regime dependent. In the investigation described, an artificial neural network approach is used to directly estimate the component fractions in gas–oil, gas–water, and gas–oil–water flows from ECT measurements. A two-dimensional finite-element electric field model of a 12-electrode ECT sensor is used to simulate ECT measurements of various flow conditions. The raw measurements are reduced to a mutually independent set using principal components analysis and used with their corresponding component fractions to train multilayer feed-forward neural networks (MLFFNNs). The trained MLFFNNs are tested with patterns consisting of unlearned ECT simulated and plant measurements. Results included in the paper have a mean absolute error of less than 1% for the estimation of various multicomponent fractions of the permittivity distribution. They are also shown to give improved component fraction estimation compared to a well known direct ECT method.

The subject of pneumatic conveying of solids is a complex one. The flow regime present in a conveying system is dependent upon: the size and shape of the particles to be conveyed, the geometry and orientation of the conveying pipe, the relative densities of the solid and the conveying air. The variable parameters present are the velocity of the conveying air and the solids mass flow rate. The variation of these two factors dictates the presence of either dilute or dense phase flow. At Manchester Metropolitan University a pneumatic conveying system, transporting polyethylene nibs, was used to investigate the implementation of a proportional and integral control system using a tomographic imaging system in the feedback loop. The aim of the investigative work was to achieve control of the air velocity and solids loading factor for the conveying system to maintain dilute phase flow at a prescribed dune level. The solids material conveyed was sensed using a PC based electrical tomographic imaging system and this was used to control the air velocity in the conveying system.

The design of sensing and control systems for twodimensional industrial processes such as paper making, plastic film extrusion and metal rolling, has drawn extensively upon concepts from two-dimensional signal processing and image processing. The performance of two-dimensional control systems is discussed in terms of spatial and dynamic bandwidths and is illustrated by examples from the control of cross-directional variations on a plastic film extrusion line. By relating these bandwidths to the specification of the required quality of the finished film and typical disturbances that enter the process, specifications are developed for the design of the actuators, the sensing system and the control algorithm. It is shown that the current generation of control systems is not suitable for controlling variations over small areas of sheet, but a system which regulates over much smaller areas could be developed using recent developments in sensors and two-dimensional control algorithms.

In the production of steel strip, the temperature distribution and cooling rates along the mill run-out table have a significant effect on the steel microstructure and hence on final material properties, e.g., yield strength, tensile strength, and ductility. Noncontacting optical temperature sensors are typically used to implement feedback control of cooling, but water spray and surface emissivity irregularities can adversely affect these sensors. Ideally, the control of cooling path should account for the progress of dynamic transformation at required points rather than the strip temperature alone. There are several reports describing the use of magnetic sensors to monitor transformation. These sensors exploit the change in the electromagnetic properties as the steel progresses through transformation, for example the austenitic phase is paramagnetic and the ferritic phase is ferromagnetic below the Curie point. Previous work has concentrated on the operation and design of individual transformation sensors. This paper now describes the use of an array of electromagnetic sensors to image the progression of transformation along a sample steel block on a pilot scale industrial mill. The paper will describe the underlying physical principles, the design of the system, and present images showing the progress of transformation along one surface of the sample.

Level monitoring instrumentation is an essential part of hydrocarbon processing facilities, and has, together with separator technology, been widely addressed over the last decade. Key issues are production capacity, product enhancement, and well-flow control. The reliability and accuracy of the level instrumentation, and its ability to monitor all the interface layers of the separator, including the thickness of the foam and the oil–water emulsion, are particularly important when considering the level instrumentation as the main sensing element in the automatic control of the separator vessel. Lately, industry focus has been placed on optimal automatic control to improve the quality of the production output, and to minimize the use of expensive and environmentally undesirable separation enhancing chemicals. Recent developments in hydrocarbon production include subsea separation stations, where the constraints placed on the reliability and accuracy of the level instrumentation are especially demanding. This paper presents level interface monitoring developments based on electrical, ultrasonic, thermal, and nucleonic physical principles for three-phase hydrocarbon separators, and introduces the notion of tomometry, meaning multipoint cross-sectional metering aiming to acquire information on the cross-sectional flow-component distribution in the process vessel intended for control purposes.

Texture is one of the main image attributes used for visual and computer-aided pattern recognition in several areas. However, texture analysis is often confounded by other neighboring or surrounding features. The auditory system has not been used much for image analysis in spite of the fact that it has several advantages over the visual system. We explore in this article the potential of auditory display procedures, including audification and sonification, for aural presentation and analysis of texture in images. We first draw an analogy between random/periodic texture and unvoiced/ voiced speech in terms of their generation based on filtering of an excitation function. An audification procedure that consists of playing in sequence the projections of the given image at several angles is then proposed. A linear prediction model is used to generate the sound signal from the projection data. Also presented is a sonification procedure to convert periodic texture to sound with the emphasis on displaying the essential features of the texture element and periodicity. In order to verify the potential of the proposed methods for aural analysis of texture, a set of pilot experiments was designed and presented to 10 subjects. The results indicate that the proposed methods can facilitate qualitative and comparative analysis of texture. The methods could lead to the use of auditory display of images as an adjunctive procedure to visualization.

In this paper, we present a new framework for the design of adaptive transforms which have overlapping and nonoverlapping basis functions. We provide the condition that the overlapping basis functions have to satisfy, and derive the optimal nonoverlapping basis functions in the energy compaction sense. We show the orientation adaptation example, where each adaptive transform is characterized by the angle of edges. In image coding experiments, the proposed adaptive transform can reduce the blocking effect because of the use of overlapping basis functions. Moreover, the use of orientation adaptation can improve visual quality around edges and lines. The performance of the adaptive transform system is superior to that of the nonadaptive transform in the rate-distortion sense. The proposed framework can be applied to any design method for adaptive transforms based on the Karhunen–Loe`ve transform.

Color images may be encoded by using a gray-scale image compression technique on each of the three color planes. Such an approach, however, does not take advantage of the correlation existing between the color planes. In this paper, a new segmentation-based lossless compression method is proposed for color images. The method exploits the correlation existing among the three color planes by treating each pixel as a vector of three components, performing region growing and difference operations using the vectors, and applying a color coordinate transformation. The method performed better than the Joint Photographic Experts Group (JPEG) standard by an average of 3.40 bits/pixel with a database including four natural color images of scenery, four images of burn wounds, and four fractal images, and it outperformed the Joint Bi-Level Image experts Group (JBIG) standard by an average of 3.01 bits/pixel. When applied to a database of 20 burn wound images, the 24 bits/pixel images were efficiently compressed to 4.79 bits/pixel, then requiring 4.16 bits/pixel less than JPEG and 5.41 bits/pixel less than JBIG.

In this paper, we present a low cost approach to design rate control algorithm for JPEG-LS, the latest JPEG lossless/near lossless image compression standard, to achieve an information loss adaptive and rate-controlled near lossless image compression. The design incorporates the major features of data compression in JPEG-LS and seamlessly adds an information loss distribution scheme into the standard to adjust the compression ratio in accordance with the variance of image content and lock the final compression ratio at a desired target specified by users. The algorithm proposed also satisfies the general requirement for industrial standards as such that: (i) the complexity of JPEG-LS is maintained at a similar level without any significant increase; (ii) the compression ratio at its near lossless compression mode is controlled at a target bit rate within the range of near lossless compression; (iii) the visual perception quality is even better than the original JPEG-LS in smooth areas; (iv) the image quality in terms of PSNR is maintained competitive to the original JPEG-LS. Extensive experiments confirm the above statements when the original JPEG-LS is used as our benchmark, which is not rate controlled.

This paper presents a combined approach for image restoration with edge-preserving regularization, subband coding, and artificial neural network. The edge information is detected from the source image as a priori knowledge to recover the details and reduce the ringing artifact of the subband coded image. The multilayer perceptron model is employed to implement the restoration of images. The main merit of the presented approach is that the neural network model is massively parallel with stronger robustness for transmission noise and parameter or structure perturbation, and it can be realized by very large scale integrated technologies for realtime applications. To evaluate the performance of the proposed approach, a comparative study with the set partitioning in hierarchical tree (SPIHT) has been made by using a set of gray-scale digital images. The experiment has shown that the proposed approach could result in considerably better performances compared with SPIHT on both objective and subjective quality for lower compression ratio subband coded image.

In this paper we study robustness of nonlinear filters for image processing by using a recently introduced method called output distributional influence function (ODIF). Unlike the traditionally used asymptotic methods, such as the influence function and the change-of-variance function, the ODIF provides information about the robustness of finite length filters used in image processing. The ODIF is not only a good theoretical analysis tool but it can also be used in real filtering situations for selecting filters behaving as desired in the presence of contamination. The applicability of the ODIF to the real image processing tasks is validated by experiments on images. We present the ODIFs for stack and L-filters which include many of the filters useful in the image processing applications. The usefulness of the ODIF in the analysis of the robustness of different filters is demonstrated in several illustrative examples by using the ODIFs for the expectation and variance.

The use of wavelets has grown enormously since their original inception in the mid-1980s. Since the wavelet data representation combines spatial, frequency, and scale information in a sparse data representation, they are very useful in a number of image processing applications. This paper discusses current work in applying wavelets to object and pattern recognition. Feature extraction methods and search algorithms for matching images are discussed. Some important issues are the search for invariant representations, similarities between existing applications and the human visual system, and the derivation of wavelets that match specific targets. Results from several existing systems and areas for future research are presented.

A tracking algorithm has been developed to efficiently track a moving object in two-dimensional image space. The algorithm employs a limited exhaustive template matching scheme that combines the accuracy of an exhaustive search with the computational efficiency of a coarse-fine template matching scheme. The overall result is an accurate, time-efficient tracking algorithm. After providing a theoretical discussion of the algorithm, three separate biomedical applications of the algorithm are described: (1) stabilizing an irradiating laser on the retinal surface for photocoagulation treatment, (2) measuring target fixation eye movement to construct pattern densities at the retina, and (3) tracking a rat swimming in a Morris Water Maze for psychophysiological studies. Results for each application is provided. The paper concludes with a discussion of the relative merits of the tracking algorithm and recommendations for methods to improve the performance of the algorithm.

Digital watermarking is a labeling technique for digital images which embeds a code into the digital data so the data are marked. Watermarking techniques previously developed deal with on-line digital data. These techniques have been developed to withstand digital attacks such as image processing, image compression and geometric transformations. However, one must also consider the readily available attack of printing and scanning. The available watermarking techniques are not reliable under printing and scanning. In fact, one must consider the availability of watermarks for printed images as well as for digital images. An important issue is to intercept and prevent forgery in printed material such as currency notes, bank checks, etc., and to track and validate sensitive and secret printed material. Watermarking in such printed material can be used not only for verification of ownership but as an indicator of date and type of transaction or date and source of the printed data. In this work we propose a method of embedding watermarks in printed images by inherently taking advantage of the printing process. The method is visually unobtrusive to the printed image, the watermark is easily extracted and is robust under reconstruction errors. The decoding algorithm is automatic given the watermarked image.

An adaptive neighborhood contrast enhancement (ANCE) technique was developed to improve the perceptibility of features in digitized mammographic images for use in breast cancer screening. The computationally intensive algorithm was implemented on a cluster of 30 COMPAQ Alpha processors using the message passing interface. The parallel implementation of the ANCE technique utilizes histogram-based image partitioning with each partition consisting of a list of gray-level values. The master processor allots one set of gray-level values to each slave processor. Each slave locates all possible seed pixels in the image with the designated gray-level values, grows a region around each pixel, enhances the contrast of the seed and any redundant seed pixels if required, and returns the results to the master. The master then sends a new set of gray-level values to the slave for processing. The subdivision of the original image based on gray-level values guarantees that slave processors do not process the same pixel, and is particularly well suited to the characteristics of the ANCE algorithm. The parallelism value of the problem is in the range of 16–20; the performance did not improve significantly when more than 16 processors were used. The performance declined when more than 20 processors were used. The result is a substantial improvement in processing time, leading to the enhancement of 4 K34 K pixel images in the range of 30–90 s.

Many image processing operations can be abstracted into matrix operations. With the help of matrix analysis, we can understand the inherent properties of the operations and thus design better algorithms. In this paper, we propose a matrix decomposition method referred to as identity-plus-row decomposition. The decomposition is particularly useful in design of parallel projection algorithms on mesh-connected computers. Projection is a frequently used process in image processing and visualization. In volume graphics, projection is used to render the essential content of a three-dimensional volume onto a two-dimensional image plane. For Radon transform, projection is used to transform the image space into a parameter space. By applying the identity-plus-row matrix decomposition method, we solve the data redistribution problem due to the irregular data access patterns present in those applications on single instruction stream, multiple data stream (SIMD) meshconnected computers, developing fast algorithms for volume rendering and Radon transform on SIMD mesh-connected computers.

Information hiding is a rapidly developing research area with a potentially enormous number of practical applications. There is already a considerable amount of commercial interest in information hiding applications. However, there is a lack of a good theoretical understanding on the subject, without which, commercial applications will face significant problems that would only hinder their success.