Design And Implementation Of CIS Image Acquisition And Preprocessing System Based On FPGA

In the modern industrial automation production process,machine vision inspection technology is more and more widely used in various fields of industrial production such as defect detection,identification and sorting.The speed and quality of image acquisition determine the inspection speed and accuracy of the product.The image acquisition system with CIS(Contact Image Sensor)as the core module has the advantages of high integration,compact structure,small size,and no geometric distortion in imaging.It is suitable for collecting surface images of printed matter,textiles and other objects.According to the requirements of the online inspection system for printed matter defects on the image subsystem,a real-time and high-speed image acquisition and preprocessing system based on FPGA(Field Programmable Gate Array)and CIS is designed and implemented.The main research contents include:(1)Aiming at the problem of image data transmission rate mismatch in different FPGA modules,an image buffer structure with FIFO + DDR2,which buffers image data through asynchronous FIFO to support fast access to image data is designed.Meanwhile,CIS and analog-to-digital conversion chip are controlled accurately based on FPGA to realize real-time and stable collection of printed image.(2)To solve the problem of image data loss in the state of large amount of transmission data and high transmission speed,a lightweight UDP protocol based on FPGA and Gigabit Ethernet is designed to realize the real-time and high-speed transmission of image data.At the same time,the error checking timeout retransmission mechanism are added into the UDP protocol,which improves the reliability of data transmission.(3)Aiming at the problem of poor image quality due to the optical characteristics of CIS,electromagnetic interference in the industrial environment,and uneven transmission of printed materials,a series of image preprocessing algorithms based on FPGA are proposed.The CIS photosensitive element cannot guarantee good consistency due to the manufacturing process,resulting in non-uniform column stripes on the image.Aiming at this sub-problem,an image gray-scale distortion correction algorithm is designed to eliminate column fringes on the image.In order to eliminate the salt and pepper noise that appears on the image due to electromagnetic interference in the industrial environment,an improved adaptive median filter algorithm is designed,which also has a better denoising effect under high-density noise conditions.Due to the uneven transmission of the printed matter,wrinkles are generated,resulting in uneven brightness on the collected images.Aiming at this sub-problem,a brightness equalization algorithm based on guided filtering and a two-dimensional gamma function is designed,which can perform adaptive brightness equalization processing on non-uniform illumination images and improve the image quality.The experimental results show that the system designed using the above scheme can realize the real-time and high-speed acquisition,transmission and preprocessing of flat printed images.When the image resolution is 300 DPI,the maximum acquisition speed can reach 128m/min,and the maximum transmission rate of Gigabit Ethernet is 975.93MB/s.After the CIS original image is preprocessed by this system,the image quality is significantly improved.The products developed based on the research results of this thesis have been successfully applied to the online defect detection systems of several printed matter manufacturers,which proves the effectiveness and stability of this system.