Research And Application Of Edge Detection Algorithms Based On Machine Vision

As the continuous progress of science and technology, the machine vision technology hasaroused widespread concern with its unique advantages. The most important feature ofmachine vision is the machine automation equipment instead of the traditional manual labor,which improving the efficiency of production automation vastly. In recent years, particularlythe substantial labor costs increase the major companies, even small and medium enterpriseson the human cost of attention, a growing number of machine vision products began toreplace the product quality inspection and supervision with quality control and supervisorystaff. Especially in harsh environments, such as sewers, high-speed high-risk environments,such in the situation as artificial vision can not be used, machine vision seems more essential.Machine vision not only greatly improve the degree of automation of the production, and thedetection accuracy, but also solves the problems of a great eye detection unsupervised,non-normative, low detection efficiency, low accuracy.Edge extraction of the image is the core of the entire machine vision algorithms, whichcontains most of the information in the image feature, provides a number of valuable featureinformation for the subsequent fitting, characterized in division, neural network classification.Extracted edge extraction is directly related to the machine vision, it is the basis of thefollow-up to a variety of image processing algorithms, edge extraction importance isself-evident, which put a lot of research. Edge extraction technology continues to explore, andso far has been the emergence of many advanced technology.Extensive literature research shows that the mainstream edge extraction algorithm contain Sobel operator, Canny operator, LOG operator, Kirsch operator, Robert operator,Robinson classical algorithm operator, etc.. Currently, the most representative of the operatorisCanny operator, which is a robust segmentation algorithm fitting the light mostly. So, Cannyoperator which use a Gaussian filter and optional dual-threshold segmentation algorithm is thefavorite of the majority of scientists, however, Canny operator also has some defect.Thispaper presents an improved based on edge detection algorithm.Based on analysis of the edge detection algorithm, the edge detection algorithmincluding Canny operator, Sobel operator, LOG operator, Kirsch operator, Robert operator,Robinson operator are firstly detailed descriped， and the corresponding algorithm and itsexperimental simulation are realized，the experimental results are summarized and analyzedafter that. After the advantages and disadvantages of various algorithms in Canny operator’sannalized，one improving algorithms is proposed and designed based on Canny’s edgedetection operator. First, a median filter is in place of the Gaussian filter and the original4x4large window is selected instead of neighborhood2x2neighborhood calculated gradientmagnitude,then direction of four differential algorithm is introducted. For the adaptive edgeextraction poor problem, an adaptive ability of dual-threshold segmentation algorithm isdesigned.In order to test the edge extraction algorithm, a automotive wheel measuring system isdesigned,and the edge detection algorithm is proposed. During the experiment, Sobleoperator, LOG operator and the proving edge detection algorithm is tested.The results showthat the diameter measured with the operator edge detection algorithm proposed in this paperhas most stable, least volatile, high precision, which achieved the expected effect.