Design And Implementation Of Product Surface Defect Detection System Based On Machine Vision

This thesis mainly introduces the design and implementation of a surface defect detection system based on machine vision.The design of the system is mainly divided into two parts: hardware platform and software platform.The hardware platform is mainly divided into two parts: cable imaging system design and computer image detection platform design.The software platform is mainly divided into three parts: software interface design,background algorithm implementation,and defect alarm implementation.The innovation of this thesis is mainly to design and implement a high frame rate and high accuracy cable defect image recognition and classification software and hardware system.At the same time,a multi-threaded segmented detection and defect type recognition and classification image analysis algorithm is proposed.,It can achieve the design goals of high frame rate defect image real-time detection and high accuracy rate classification at the same time.In the hardware platform design,the hardware selection of the detection system is mainly introduced,including the selection of industrial cameras with multiple pixel imaging materials,the selection of lenses with multiple optical imaging principles,the selection of light sources with multiple imaging effects,and multiple computing capabilities.The selection of central processing unit and image processor,etc.At the same time,the imaging system design method is introduced and the physical connection diagram is shown.In the software platform design,the interface design and implementation of image display,defect detection,defect alarm,defect report generation and other functions are mainly introduced.At the same time,the design and implementation of multi-threaded image transmission,storage,processing,defect alarm and other functions are introduced.In the background algorithm implementation,the specific implementation method of multi-threaded segmented detection and identification of defect types is introduced.The detection algorithm mainly uses the threshold segmentation and morphological processing in the traditional digital image processing,and the median filtering and downsampling algorithms are mainly used in the image preprocessing operation.In the algorithm for identifying defect types,the residual network deep learning model constructed by the convolutional neural network is mainly used,and the continuous optimization of the model is realized through the back propagation algorithm.At the same time,it introduces the realization of high-efficiency and concurrent processing of image algorithms using computer multi-threaded collaborative computing.Finally,through the analysis of on-site test results of real-time cable production data,it is concluded that the algorithm is more effective and the hardware platform is more stable.