Research On Surface Defect Detection Algorithm Of Workpiece Based On Machine Vision

In recent years machine vision technology is a research hotspot in the world of science, and many of the cutting-edge technology became realized and optimized on the basis of it, such as self-driving car, tumor recognition and so on. Also machine vision technology is widely used in the field of industrial detection, and particularly in the detection of the defects of complex work piece, the machine vision technology has a great application space.This paper built a machine vision measurement platform for work piece inspection, high precision screw bolts are selected as the object of research, the surface defects and geometric defects of the bolts are detected, the speed and precision of detection are both higher than artificial detection. The detection method is able to meet the requirement of production line testing.For the surface defects detection of the bolt, different types of defects lead to different texture changes. Adopt the method of analyzing texture characteristics of the bolt thread, research on the characteristic parameters which influence the texture. Use the characteristic parameters of standard bolt as the template to build characteristic daTab.ase for defect detection test. The successful detection rate reach 98 percent, the detection speed is 4.5 per second.For the geometry defects detection of the bolt, this paper mainly study at the size detection, such as pitch, major diameter, minor diameter and thread height. By extracting the edge of thread, using the improved Harris operator to detect sub-pixel corners, then, calculate the thread geometric parameters. Use the standard geometric parameters as the template for threshold detection. The detecting accuracy reach 0.2mm, the speed of detection is 3 per second.The machine vision measurement platform for work piece inspection that designed in this paper has good detecting effect on the work pieces with complex texture and structure, the whole detection process takes 7.5 seconds, and can meet the requirements of industrial automation detection.