Image Acquisition And Preprocession System For Vehicle Surface Defects Automatic Defection

Automobile surface is prone to various appearance defects and needs to be detected in time in the process of production and transportation.At present,the visual inspection is basically rely on manual visual inspection.Therefore,an automatic and intelligent detection system is urgently needed.Machine vision inspection is an emerging inspection technology that provides guidance for the detection of defects in the appearance of vehicles.In this paper,an image acquisition and preprocessing system for vehicle appearance defects automatic defection is proposed to provide a system platform,image detection preprocessing technology and defect detection method.The main work is as follows:Firstly,based on the analysis of technical requirements of the system and the analysis of technical routes,the overall design of the vehicle image acquisition system is performed,the hardware of each system is selected,and the functions and modules of the system software are divided.Secondly,the image preprocessing algorithm is studied,including three aspects.In the aspect of image denoising,the BM3 D algorithm is analyzed in detail.Then the gray mean value method,the comparison result recording method and the wavelet domain denoising method are effectively fused together to achieve the improvement of the BM3 D algorithm and reduce the computational complexity of the algorithm.In the aspect of image registration,based on the SURF image registration,the new DAISY feature descriptor is substituted for the original SURF feature descriptor,and the signature matching random kd tree is used to complete the feature matching.An image registration method based on the improved SURF_DAISY algorithm and the verification random kd tree is proposed.In the aspect of image edge detection,the edge detection method based on modulus maxima method of wavelet transform and the edge detection method based on new morphology are analyzed.Finally,the two algorithms are combined and the paper puts forward a noisy image edge detection method which combines modulus maximum method of wavelet transform with improved mathematical morpholopy.At last,the system hardware is builded,the system software is programmed and some of the vehicle body defect detection tests are performed.The test results verify the feasibility of the system hardware and software and the effectiveness of the image preprocessing algorithm.The entire system has reached the technical requirements.The system provides a good system platform and technical support for the following defects automatic detection system.