The Spatial And Frequency Domain Analysis Methods For Surface Inspection Of Industiral Products Based On Machine Vision And Related System Implementation

Since online surface inspection based on machine vision has been an effective way to upgrade the surface quality of industrial products,it's widely used in many fields of industry.However,the accuracy and precision of the traditional surface inspection method still need to be improved.In the field of surface inspection based on machine vision,spatial and frequency domain analysis is an effective method for image processing.As its accuracy and precision are better than the traditional image processing method in image segamentation,image registration and feature extraction,it is getting more and more attention of researchers.According to the theory of machine vision,a distributed system for online surface inspection of industrial products is presented.The system is mainly based on the techniques of spatial and frequency domain analysis,in which the characteristics of surface image are also concerned seriously.Both the algorithms and the techniques to implement the inspection system have been studied in depth.In the application of wide-format and high-precision surface inspection,as the traditional way to construct the system with a single camera and a single computer can hardly fulfill the real-time requirement of online inspection,a distributed system is proposed,which can satisfy the real-time demand of online surface inspection effectively.Using the image preprocessing algorithm,the disturbances can be eliminated from the image,and the stability of subsequent algorithms can be improved effectively.According to the characteristics of surface image,an adaptive median filter for vision-based industrial products inspection is presented to filter the salt and pepper noise;a nonlinear gray transformation based on quadratic polynomial is adopted to improve the situation of inconsistent image brightness at different moments;and an adaptive threshold surface algorithm is put forward to reduce the effect of uneven illumination.The spatial and frequency domain analysis method can be used to filter out the texture from the surface image,which can help to improve the precision and performance of the inspection system.In this dissertation,a Fourier analysis method for fringe pattern segmentation is put forward to depart the defect from the fringe background.And the method can be improved to fulfill the requirement of online ream inspection of float glass.Meanwhile,a threshold surface method for image segmentation based on wavelet is presented,with which the unstable texture can be filtered out from the background effectively.According to the multi-resolution techniques related to spatial and frequency domain analysis,an image registration using edge features based on wavelet is provided to determine the position of defect conveniently.In the vision-based inspection process of industrial products,the intra-class variance of defects is large,and sometimes the inter-class variance of defects is not obvious.So an enhanced feature extraction method based on wavelet is presented to describe the defect features with the gray features,the geometric features,the multi-scale local texture features,and the multi-scale global texture features.Since the gray features and the geometric features are used to describe the intra-class variance of defects,the multi-scale local texture features are adopted to describe the inter-class variance of defects,and the multi-scale global texture features are employed to reduce the effects of noises,the stability of the system can be ensured by the method effectively.According to the principle to recognize an object,a general purpose classifier for the defect classification of industrial products is presented.Since the enhanced defect features are used in the classifier,the adaptivity of surface inspection system can be improved,and the development cycle can also be cut down at the same time.As the classifier introduces the mechanism of evolution,the accuracy can be enhanced along with the accumulation of uses.Finally,an improved online defect detection system for float glass using the STFT method,and an online surface inspection system for steel strip based on wavelet analysis are implemented according to the methods mentioned above,which can fulfill the requirements of surface inspection in practice.