Research On Surface Defect Detection Technique Of Screw Thread Based On Image Processing Technology

Screw threads are important connection, transmission and tighten components,which are widely used in machinery, automotive, transportation and other majorindustries. Surface defects of screw threads directly affect the performance of threads,resulting in a variety of security risks. For the surface has defect of the screw thread, itmust immediately be removed to avoid unanticipated consequences. In the domestic,surface defect detection of screw thread is still stuck in the artificial visual detectionstage. It not only spends a lot of manpower, but also prones to false alarm and misseddetection, which has greatly limited the improvement of production efficiency andproduct quality. Therefore, carrying out the topic research has important practical valueand social benefits.Having considered various detection methods on surface defect detection of screwthread in and abroad, machine vision inspection method is selected as the topic researchmethod in this dissertation, which based on image processing. Firstly, the screw threadsurface image acquisition experimental platform is designed on the base of analyzingindustrial cameras, optical lens and light source and lighting. Secondly, the classicalalgorithms for the detection of surface defects are researched and compared, such asdiscrete Fourier transform (DFT), discrete cosine transform (DCT), discrete wavelettransform (DWT), and principal component analysis (PCA). The results show that thefour methods are good, expect DWT-based approach cannot detect the defects thatparallel to the texture of the screw thread. DCT-based method shows more efficient incomputation than other three methods due to it all deals with real number and has fastDCT algorithm. Based on this reason, an image enhancement process based on DCT isused in this dissertation. Finally, the screw thread surface defects detection algorithmbased on image processing is designed and implemented. DCT-based method is used toachieve defect image enhancement and the statistical process control (SPC) binarizationmethod is used to achieve defect segmentation. In segmentation post-processing, thedefects are detected by using connected component labeling operator with8-adjacentwhich will have the noise blobs with small areas be removed. In order to achieve theautomatic determination of the detection algorithm parameters, the automatic selectionmethod of high power threshold k1and control factor k2value are given in thisdissertation. Experiments prove that the defects detection results are good with the screw threadsurface defects detection algorithm based on image processing in this dissertation. Thedefects both on the crest and root of the screw thread can be detected, so the purpose ofthe screw thread surface defect detection is achieved. The defect segementation resultcan be used in the defect recognition as it preserves the defect shape. It provides animportant basis for the realization of the automated management of the quality of thescrew thread.