Tool Wear Detection Technology And System Based On Computer Vision

With the rapid development of industry intelligence, computer technology iswidely used in machinery manufacturing industry in which to improve theproductivity and save cost. Especially the rapid development of the aerospace,automotive and other high-tech industries, the quality of modern manufacturing havehigher requirement, and the tool wear in the metal cutting process can stronglyinfluence the machining accuracy and surface quality of the workpiece, so the toolwear condition monitoring technology has became an urgent problem. Real-timeinformations of tool wear can be effectively obtained by tool wear conditionmonitoring technology, to determine to automatic change tool if tool has damaged.Or the relationship between the amount of tool wear and processing parameters isquantitative researched, to optimize cutting parameters, after know the tool wearcondition, it can guarantee the machining quality, improve production efficiency,retard tool wear, extend tool life,and save the manufacturing cost.In this paper, focuses on research of based on computer vision tool weardetection technology and systems. According with tool worn surface imagingprinciple, the tool wear image pixel gray value distribution regulations is summed up.Explores the image processing technology in the application of the algorithm to detecttool wear, experimental verification image processing tool wear optimal algorithm,and proposed a series of improvements and new algorithms based on this,thendeveloped algorithm flow of measurement. First, two image enhancement technologyare applied, median filtering and contrast stretching, for image enhancement, and theautomatic determination algorithms of the upper-and lower-thresholds arerespectively presented according to the Otsu method and B-spline curve fittingmethod. The gray-contrast between the wear area and the background area can beaccurately enhanced as well as the gray-contrast between the wear area and theunworn areas. Secondly, according to the stable region within three areas and thenon-stable region at two edges in the tool wear image, the local gray-level variance issuggested for the boundary extraction. The adaptive threshold of the local variance isagain defined to segregate the tool wear region from the tool wear image clearly. On this basis, the tow kind of morphological methods, closing and filling holes, can beemployed to fill out the holes of the segregated part so that the correspondinggeometric parameters of the wear areas will be precisely achieved, after the tool wearis extracted feature and calibrated dimension. Then analysis the measurement data andcompares measurement error of segmentation algorithms presented in this paper. Theresults show that the local variance threshold segmentation algorithm proposed in thispaper have an absolute error less then1.325%.Finally, a tool wear measurement experiment system is designed as a graphicaluser interfaces(GUI) in Matlab, and tool wear geometric parameters measured by thesystem. Experimental results show that the designed system is characteristic of highlevel automation and simple operation, In addition, verily that algorithms proposed inthis article can completely and clearly extract boundary, and have high-precisionmeasurements and strong anti-interference ability.