Research On Variable-step Sampling Mechanism Used In Fast And Automatic Recognition Of Products

With the improvement of industrial production technology and requirements of productquality, requirements for circumferential detection in internal part or surface of vital product,which are required to mass-produced in the automotive industry, aerospace, defense and otherfields. However, the current detection methods is still given priority to artificial detection,fixed step method is used to test the objective of specific products in the current detectionsystem, which has been applied in a few research field. Because the determine of a fixed stepmostly depends on the minimum structure size in multiple identify the target of the product,Although using the fixed step to assure the recognition of the minimum structure size,enormous data will result if having a over-sampling on other targets, both reduce therecognition accuracy and fail to meet the real-time problem of on-line test.In fact, there are different sizes of inspected target in the internal part of products in theprocess of recognition. The common representative black and white pair in the visual field isused in the research on the system of variable step sampling which is limited in structure sizeand space distortion. Firstly, the function relationship is formed between sampling step andspace structure size represented by different line pairs, and then function relationship isformed based on sampling step and space structure distortion that is showed by different slopeof line pairs. The experiment result shows that sampling step is increased in minimumstructure size, and the relationship curve of space distortion satisfied the sinusoidal law.The sampling method of variable step-size is applied in X-ray automatic detection fields,six typical sites in inspected products are defected in this paper. First, the samplingmechanism of variable step and the known a priori least structure size is used to determine thesampling step of the inspected target; Second, find solution in the standard library in allrandomized inspected products image with the gray level statistic trend analysis method;Finally, the advantages and disadvantages of parallel detection and serial detection arecompared in the process of accuracy detection. Experiments show that on the premise of effective and accurate, the method of gray statistical trend analysis takes less time than theimage matching method of in finding the optimal place in the random inspected products fromthe standard library. In the process of automated inspection of interior assembly structures ofproducts, parallel detection method used in this paper without classifying defects have highlyefficient than serial detection. The sampling method of variable step is applied in both theX-ray defect detection of internal structure and visible light filed of the surface defectdetection.