Design And Implementation Of Image Processing System For Smart Industrial Camera Based On Multi-processor Co-design

Under the general trend of automation and intelligence in the industrial manufacturing industry,machine vision inspection technology replaces traditional artificial vision detection in the detection process with high efficiency and high quality.Machine vision inspection has the advantages of stable,non-contact,real-time online,etc.,which can effectively help the manufacturing industry to upgrade.At present,in the DA process of wafer fabrication,the inspection work is still done manually.Therefore,machine vision inspection technology is used in DA process to develop the image processing system of intelligent industrial camera that can meet detection requirements in this process,and realizes the automation of detection.The main research and development work of the project includes the follows:(1)After investigating the research status and development trend of smart industrial cameras,analyzing the advantages and disadvantages of smart industrial cameras based on different hardware platforms,and taking into account the characteristics of the inspection requirements of the DA process in wafer production,Zynq-7000 SoC is chosen as the hardware platform to do system development.(2)Based on the multi-processor architecture of Zynq-7000 SoC,using the software and hardware coordination idea,the system design has been proposed and system functions have been completed through co-processing with PL and PS.(3)Utilizing Vivado HLS development tool to design image processing algorithms that can be synthesized,including the illumination compensation algorithm,the OTSU binarization algorithm,the Hough transform line detection,the Hough transform circle detection and color segmentation.After simulation,these algorithms have been packaged into IP cores and integrated into the system hardware to implement the image processing function of the PL side.(4)Embedded Linux operating system has been built on Zynq platform,and QT,OpenCV and ZBar also have been transplanted.The OV5640-based image acquisition display function has been implemented both on bare-metal and embedded Linux system respectively,the hardware construction and image data flow have been introduced,and the related driver configurations in Linux have also been completed.(5)In embedded Linux system,the processed image that via the PL side has been further processed to implement functions such as Barcode/QR code recognition.Finally,association test has been carried out on the Zynq platform.In wafer DA process,the functions of the system include: the detection of blue and white membrane,the detection of presence or absence for barcode,the identification of barcode and other functions.