Detection And Classification Of Laser Level Conical Mirror Defects

Laser level is widely used in Windows and doors installation,floor laying,hanging decoration installation and mason maintenance and other fields.Laser level is to emit laser from the bottom up through the laser,laser irradiation on the cone mirror,due to the role of diffuse reflection can be laser light clear irradiation on the wall,cone surface scratch,not on the car,dirty area and other problems will directly affect the laser light,so the quality of the cone mirror is particularly important.At present,the laser level cone-mirror defect detection mostly adopts manual method to detect the defect.As the long-term manual detection increases the degree of eye damage,the attention can not be focused all the time during the detection,so the detection accuracy and detection speed can not be guaranteed.Based on the above problems,this paper implements the long distance and short distance defect detection algorithms of cone mirror and classifies the short distance defects.This paper first analyzes the defects of the conical mirror laser level principle,in view of the laser light at a distance short floating problem and close the laser light show eight lines,flare and miscellaneous line defects such as the scope of the problem,the design of two kinds of image acquisition,a camera located 5 meters distance detection,another camera distance tests a 30 cm position to capture.The laser interference fringe images collected from a long distance are mainly for the information of the main line and the secondary line.Based on this,an improved algorithm of Zernike moment method + iterative method to select the best threshold value is proposed.After the two algorithms are used to extract the edges of the defect image,it is found that the improved Zernike moment method can extract the edges of the main line and the secondary line more accurately.Finally,by maximum entropy value method + genetic algorithm to close four defects threshold segmentation,can be miscellaneous line,eight line,not the car,defects such as scratches,partition,the image center of horizontal ordinate,area,roundness,grayscale average and gray entropy features parameters setting of defects,close to classify four kinds of defects,Based on this,the maximum fitness parameter of genetic algorithm was used to optimize the penalty parameter and radial basis kernel function parameter of support vector machine(SVM).The experimental analysis of classification accuracy of the two methods showed that SVM based on genetic algorithm parameter optimization had higher classification accuracy.