Design And Implementation Of Pavement Crack Detection System Based On FPGA

With the rapid development of highway construction and gradual improvement of road network construction in China, road maintenance work has been paid more and more attention. Pavement crack is the main form of road diseases. It is also an important indicator of the road quality assessment. The traditional manual detection and recognition methods are not only affected by the subjective consciousness of highway maintenance workers, but also waste a lot of time, so a research for pavement cracks automatic detection and recognition is particularly urgent. The system based on FPGA (Field Programmable Gate Array) can meet the needs of high speed image processing with its hardware features. What’s more, SOPC (System on a Programmable Chip) technology can make the design more flexible, and achieve software and hardware in-system programming and update.This thesis designs a pavement crack detection system based on FPGA. The system uses Altera DE2 development board as the hardware platform, a CMOS (Complementary Metal Oxide Semiconductor) image sensor to collect pavement crack materials, and a VGA (Video Graphic Array) device to display images collected and processed. When the corresponding toggle switches are snapped, the system will output the result via a LCD.The whole system consists of four modules, including image capture module, image cache module, image display module and crack detection module. It adopts a synergistic manner through software and hardware. The former three modules are written in Verilog hardware description language. The crack detection module is designed by SOPC Builder and programmed in C language in Nios II IDE. Crack image processing is divided into three parts, they are preprocessing, feature extraction and classification identification. The thesis discusses and analyzes the selection of algorithms in each part. In the preprocessing part, the system transfers the RGB level image to the gray level image firstly. Secondly, it uses the piecewise linear transformation to enhance the image. In this thesis, the interval boundary points are automatically computed based on iteration method. Thirdly, it adopts the improved adaptive median filter to smooth the image. Then it uses Sobel operator in four directions to detect edge and uses binary morphological operation to inpaint the image. Finally, it extracts the crack skeleton. In the feature extraction part, according to various types of characteristics analysis of pavement cracks, three kinds of features are extracted. They are projection features, features based on Proximity algorithm and features based on density factors of distress. In the classification identification part, classification is completed based on BP neural network. The density factors of distress can get the best results. Finally, the system calculates the geometric characteristics of cracks and evaluates the damage degree. After debugging and improving repeatedly, the system can realize the functions of pavement crack detection well.