Design Of Multi-Channel CMOS Image Acquisition System And Its Application In Glass Defect Detection System

With the improvement of industrial automation,the yield of glass increases continuously in unit time.In order to meet the requirements of users for product quality,the manufacturing enterprises widely use the real-time detection system of glass defects based on machine vision to improve production efficiency and ensure product quality.The development of glass defect detection system based on ZYNQ SoC processor is of great significance for improving flexibility,solving the problem of high data throughput and reducing system cost.In this paper,for the key technologies involved in the glass defect detection system based on the ZYNQ SoC,firstly,the control circuit of the array CIS and linear CIS is introduced and verified by simulation;Secondly,for the color space conversion and the image preprocessing algorithm,the hardware implementation use pipeline method is presented,which is verified by the joint simulation of Matlab and Modelsim.Then,the bus interface of array CIS and Linear CIS is designed and integrated into LEON3 SoC and Zynq-7000 SoC respectively.Finally,according to the ZYNQ SoC scheme proposed in this paper,a prototype system for large-scale glass defect detection is designed,and the function verification of glass defect image acquisition and processing is completed.The main work of this paper is reflected in:(1)Design a general control IP core suitable for array CIS of DVP interface,including SCCB controller compatible with 8-bit and 16-bit register address width,DVP video acquisition controller and DMA controller supporting single,INCR,INCR4,INCR8,INCR16 burst transmission.(2)Design a IP core for multi-channel linear CIS image acquisition,including the AXI4-Lite interface for configuration and the AXI4-Stream interface for data transmission,which reduces data transmission delay and improves data throughput.(3)Using the pipeline design method to design the module of RGB and YCbCr color space conversion and the image preprocessing algorithm including image filter,image enhancement,image edge extraction and binary morphological filter,which improves the speed of image parallel processing.In addition,based on the zedboard platform,a prototype system for large-scale glass defect detection is designed and verified.On the software level,the NEON coprocessor acceleration and OpenMP parallel programming method are used to improve the efficiency of data handling and processing.Six channel image acquisition can be carried out at the same time,and the detection accuracy reaches 0.09mm,which preliminarily verifies the feasibility of the whole system scheme.