Research On Image Processing In Cable Surface Defect Detection System Based On Zynq

Cable is a modern industrial production material, and its surface quality affects the performance of the final product. The traditional artificial visual inspection methods are low reliability, poor stability and low efficiency, which can not meet the real-time inspection requirement. With the rapid development of computer technology and digital image processing technology, machine vision inspection technology has made considerable progress. Instead of the traditional artificial visual inspection, by using the cable surface defects detection based on machine vision, high-speed and real-time online inspection could be realized and the cable quality could be increased.The most important part of the machine vision system is the image processing. The difficulty of this image processing system is real-time detection of high-speed movement of the cable with high-accuracy and low false detection rate. It is hard to complete the whole image processing system only using ARM or FPGA. Therefore, cable surface defect detection vision image processing system based on Zynq is designed in this dissertation. Zynq is a new platform which was released by Xilinx in 2010. FPGA and dual-core Cortex-A9 processor are integrated together in Zynq. Zynq is different from the FPGA which embeds processor. The processor is independent of FPGA and can configure FPGA. The development process is processor-centric. Hardware/software co-design technology is used to greatly improve the processing speed of the system. Wide-bandwidth communications between ARM and FPGA is achieved using AXI bus technology. The contents are divided into four parts shown as follows.Firstly, the image processing algorithms of the cable surface defect detection is designed. Differential algorithm is used to remove the negative impact of undesirable background. Cable region is extracted by using connected component labeling algorithm. Width characteristic and projection characteristic of cable which is calculated by polynomial fitting method is used to inspect the defect. The experimental results show that the image processing algorithm presented in this dissertation can inspect the shape defect, scratches defect, damaged defect and hole defects. The minimum detectable defect size is 0.8 mm and the inspection accuracy rate is over 90%.Secondly, the overall scheme of image processing system based on Zynq is designed. The problem of low processing speed of general-purpose processor is solved and the real-time detection of cable defects is realized. In order to increase the processing speed, the running time for each step of the image processing algorithm is analyzed. Some algorithms which consume long time are implemented on FPGA part of Zynq, and the other algorithms are implemented on ARM of Zynq. Wide-bandwidth communications between processing modules is realized using AXI bus technology and DDR memory chip. The experimental results show that the processing time is reduced 86.9% by using the hardware/software co-design method compared with software design method.Thirdly, the hardware part of Zynq is designed. Image processing module based on FPGA is designed using parallel processing technology and pipelining technology. And the entire IP-based hardware platform is built. The hardware modules complete a pixel processing within one clock cycle, which achieves the optimal level of image processing based on FPGA. Theoretical analysis and simulation results show that the 1600×1200 image processing speed of hardware module reaches 100 fps when the system clock is 200 MHz.Finally, the software part of Zynq is designed and the overall system has been tested and verified. Software graphical interface based on Qt and software image processing algorithms has been designed. Embedded platform has been built on Zynq for co-test of hardware and software. The experimental results of the simulating cable production test platform show that the defects on the surface of cable which is 6 mm diameter and moves at the speed of 1 m/s can be inspected by cable surface defect detection system based on Zynq. The performance of our system reaches the advanced level of the similar products abroad. To our knowledge, cable surface defect detection system based on Zynq is the first report in China.