Detection Technology Of Weak Defects On The Surfaces With Different Property Based On Machine Vision

The development of high precision machining and detection technology in defense application research and the development of intelligent manufacturing technology in industrial automation,are closely related to "machine vision detection technology".Compared with traditional manual visual inspection,machine vision detection technology,with standardized processes,can overcome the subjective arbitrariness of artificial detection,and can avoid secondary damage to the detected components by adopting the way of non-contact.Therefore,machine vision is the optimal method of surfaces defect detection and is widely used.However,with the improvement of processing technology,the improvement of surfaces quality requirements and the development of product diversification,the sensitivity and the anti-interference ability of the corresponding machine vision detection should also be improved during the test.For example,the detection ability of small defects and the anti-interference ability of surfaces complex noise,background and complex texture in the machine vision system are put forward with strict requirements,in the detection of weak defects on the surfaces with different properties.To solve the problems of low detection capability and low detection efficiency of weak defects on the surfaces with different properties,three representative sample surfaces have been discussed in this paper.And different machine vision detection methods and detection algorithms have been studied,based on the different properties.The first type of surface to be detected is the super smooth optical surfaces(SSOS),both of which are super-smooth surfaces after polished.The second type of surface to be detected is the single sided polishing optical surfaces(SSPOS),one of which is the super-smooth surface and the other of which is the matte surfaces.The third type of surface to be detected is the metal arc surface with complex texture(CTMAS),which consist of the matte surfaces with particle details and the arc surface.That is super smooth optical surfaces(SSOS),single sided polishing optical surfaces(SSPOS),and metal arc surface with complex texture(CTMAS).On the super smooth optical surfaces(SSOS),such as the high power neodymium glass surfaces,scratch defect is the main defect control object.The depth of the common scratches observed usually,is among 200nm～400nm.But there are also some weak scratches below 50nm.Because of the existence of weak scratches,it tends to have a bad or even fatal effect on the high power laser operation.However,in the detection process,the scattered light intensity generated by the weak scratches is about 20%～50%as strong as the scattered light generated by common scratches with the same width.Therefore,this kind of weak scratches can be easily undetected in the visual inspection of the human eyes and the machine vision system.To solve the undetected problem of weak scratch on SSOS,a new detection technology has been proposed including a microscopic dark-field scattering imaging method and the algorithm of adaptive smoothing and morphological differencing(ASMD).According to the different effects of weak defect and background on incident light,the ability to detect weak scratch in the system has been improved by selecting appropriate light incident angles and designing a dark field imaging method.However,the contrast between the defect and the surrounding background is still weak in the detection image,and it is easy to be disturbed by background fluctuation and image noise in the segmentation process.So,based on the differences in spatial domain and morphology of weak scratches,noises and backgrounds,the algorithm of adaptive smoothing and morphological differencing(ASMD)has been proposed in this paper.On one hand,the noise can be smoothed and background can be eliminated to an extreme.On the other hand,it can avoid the weak defect information being excessively smoothed,so that the weak defect information can be segmented from the complex image.On the single sided polishing optical surfaces(SSPOS),such as sapphire substrate with the single side polished surfaces,one of which is the super-smooth surfaces and the other of which is the matte surfaces,the defects that affect the surfaces quality are mainly the scratches defects.These scratches can increase the probability of dislocation of the epitaxial layer during the growth process of the substrate surfaces material,and increase the stress between the wafer and the epitaxial layer.However,due to the diffuse reflection effect of the matte surfaces on the incident light,the weak defect will be covered by diffuse reflection light during the detection,resulting in a missed inspection of the machine vision system.To solve the undetected problem of weak defect on SSPOS,a new detection technology has been proposed including coaxial-incident and telecentric bright field imaging method(CITBFI),and the algorithm of visual differential excitation and double discrete Fourier transform(VDEDDFT).The different effects of the top and the bottom surfaces on different types of incident lights,and the different reception effects of different lenses on the ligh,have been analyzed.Then a telecentric bright field imaging detection system with coaxial parallel incident light is designed.As a result,a certain amount of gray variation has been occurred in the defect location of the test image.However,the detection is still affected by the diffuse reflection of the matte surfaces.As a result,the gray level of the defect location is weak,and the contrast betweem the defect and surrounding noise,texture,uneven background is low.Further,according to the visual characteristics of human eyes and the characteristics of different texture information in different fields,a weak defect information extraction algorithm is designed,based on visual differential excitation and double discrete Fourier transform(VDEDDFT).Therefore,the information which contains only the defect texture can be divided from the complex image.On the metal arc surface with complex texture(CTMAS),the surfaces consist of the matte surfaces with particle details and the arc surface,wich will reflect different directions in the detection light of the machine vision system.And the image of the weak defect in the detection system is easily concealed by the complex background light image.Therefore,the defects may be undetected,and the surfaces appearance quality of the product will be affected in the subsequent processing.To solve the undetected problem of weak defect on CTMAS,a new detection technology has been proposed including multiple angle incident and telecentric bright field imaging method(MAITBFI),and the algorithm of wavelet correlation and gradient similarity growth(WCGSG).According to the characteristics of reflected light generated by the arc surface and matte texture,telecentric bright field imaging system with multiple angle incident has been designed,and the grayscale jump in the defect location is retained in the test image.However,the complex texture and arc surface will produce larger fluctuation of gray scale,so as to make the contrast of defect texture and the background texture very weak.And the The grayscale jump of defect location may be weaker than that of detail texture.So,according to different characteristics.of different textures,background and weak defects in different image transform domain and different scales of transform,the algorithm of wavelet correlation and gradient similarity growth(WCGSG)has been proposed.As a result,the complex background information can be eliminated and the weak defect information can also be enhanced.This paper has set up three different machine vision system,based on the detection technology of weak defect on the surfaces with different properties,which have been applied to the automatic detection of the defects on the Neodymium glass surfaces in the Shenguang III system and the automatic detection of the defects on the surfaces of industrial products.Besides,the detection technology of weak defects on the surfaces with different property can lay a foundation for the development of ultra-precision machining in defense applications and the development of intelligent manufacturing in industrial applications...