Research on Recognition Algorithm of Tunnel Leakage Based on Image Processing 2020-01-5133

With the fast development of tunnel engineering in China, the problem of tunnel engineering diseases has become increasingly serious. Water leakage is one of the main diseases of the tunnel, at the same time, it is the origin of other diseases of the tunnel. According to the smart highway project, this paper suggests an image processing-based tunnel leakage recognition algorithm aiming at the characteristics of many tunnels and a serious environment. Through the research of the learning vector quantization (LVQ) neural network and the Canny edge detection algorithm, and the neural network was optimized in combination with the genetic algorithm, the severity of water leakage disease is recognized. In conjunction with intelligent data processing ways, the paper solves the current problems of high cost, high cost and low detection efficiency in tunnel leakage detection, increased the speed of information feedback, as well as provided new methods of guarantee for tunnel operation safety. Using the advantages of the LVQ algorithm in the field of identification to identify water leakage, its identification accuracy is higher than traditional neural network algorithms. Through 1500 iterations, the average recognition accuracy is over 95%, and the false detection rate is 2.87%, the missed detection rate is 2.13%, as well as the time consumption is less than 20ms. After identifying the water leakage, perform grayscale binary processing, expansion and corrosion denoising. Then Canny edge detection is performed on the image to extract the leakage contour, and the growth rate of its contour area is used to judge the leakage severity.

Through practical application and experimental results analysis, it is evident that the algorithm has good usability and robustness. In the later stage, software detection and hardware detection can be combined to avoid serious errors caused by transmission failure or equipment failure of certain information. More comprehensive and accurate reasoning with judgment will be obtained than any single source information.