Roads are important parts of infrastructure. The detection of road condition plays an important role for the traffic safety. Vehicles, weather and other factors will cause different types of damage to the road surface. To avoid this happening, the commonly used method is manual inventory at present, which is time-consuming, laborious and easily leads to omissions. In order to overcome these shortcomings, this paper presents a method of road damage detection based on machine vision, which is more efficient and relatively cheap. To realize the method, the author used the Raspberry Pi, acceleration sensor, GPS module, Neural Compute Stick and camera to complete the design of intelligent inspection terminal. Then the author investigated the common types of road damage, including long strip cracks, reticulation cracks, potholes, and rutting. After that, an SSD-mobilenet architecture was modified and a database including a large number of images for different types of damage was built. The SSD-mobilenet was trained and validated with the built database. Transplanting the SSD-mobilenet to the intelligent inspection terminal, which could realize the road damage detection based on machine vision. The result shows 80.87% average precision (AP) ratings for different types of damage and proves the proposed method is effective.
A safe and healthy road condition plays a supporting role in the public travel and the national economy. Therefore, effective management and maintenance methods have become the key problems that the researchers and engineers are urgently solving, early damage detection and warning are also important for disaster emergency treatment, but some traditional road damage identification methods are often costly and need to be equipped with professional persons. Due to the complexity of pavement conditions, some existing defects datasets are not perfect, although the accuracy is high, they cannot be put into practical use. Based on the object detection technology of deep learning, the author introduced a novel method which is more effective and relatively cheap. In this paper, 5966 images with road damage of different angles and distances were collected, and the damage categories included Lateral Crack, Longitudinal Crack, Pothole and separation, Alligator Crack, and Damage around the well cover which had never been considered in the datasets in any researches. After training with GPU using convolutional neural network, the average precision can reach 96.3%.
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.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
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.