Surface Defect Detection Technology Based On C66x-DSP

With the development of society and the improvement of living standards,people are increasingly demanding quality of life.Using machine vision to improve the quality of products and life is an important way to apply and realize artificial intelligence systems.Among them,digital signal processor DSP,with its high-speed real-time processing and software programmable features,has become an optimization solution for engineers to study embedded systems in the fields of digital image processing,speech processing and radar.TI's multi-core processor,TMS320C66 x series DSP,is widely used in various fields due to its powerful processing capabilities.In recent years,with the emphasis on the quality of paper surface,more and more household papers have been required to be clean and flawless.However,the manual detection method of paper production industry has not met the speed and precision of the production line.The detection of paper surface defects based on machine vision have attracted the attention of scholars which are domestic and overseas.This paper has important research value based on embedded DSP platform to implement surface defect detection.This system mainly studies three aspects of paper surface defect detection algorithm,DSP multi-core system development design and PC software development?1.Firstly,the algorithm of surface defect detection is deeply studied in this paper.A surface defect detection method based on relative uniformity between classes is completed.Assuming that the defect foreground has gray uniformity,a threshold model with relative homogeneity between classes is established by combining the method of maximum inter-class variance.Assuming that the background does not have gray uniformity,the optimal threshold is obtained by using information entropy combined with the threshold model.In order to better verify the effectiveness of this defect detection algorithm,several classical threshold segmentation methods and the improved Otsu threshold segmentation method proposed in recent years are compared.2.Secondly,In-depth understanding of the KeyStone multi-core SoC architecture developed by TI,embedded SYS/BIOS real-time operating system,Ethernet communication module,multi-core development method,multi-core processing model and inter-core communication mode of multi-core DSP.Using an operating system-based master-slave model,and the Shared Region and Notify methods for inter-core communication.On this basis,the paper disease image detection algorithm is transplanted to realize the development of multi-core system.The program is optimized by rapid speed caching,inlining function and compiler optimization options,so as to improve the system performance.3.Finally,in order to verify and view the detection effect,a QT-based monitoring system is designed on the PC side.The PC sends the images to be detected by the DSP,and the DSP receives the images through the network and detects the the image,and then transmits the binary images detected by the DSP to the PC,and uses the PC to determine whether there is a defect.The surface defect detection system based on C6657-DSP designed in this paper can guarantee the accuracy and has certain engineering practical value.