Automobile Wheel Recognition System For Flexible Automatic Product Line

With the output of automobile wheels goes up rapidly these years, too many disadvantages come up when the various wheels in the produce lines are recognized by the manpower. For example, too much workload will burdened on the workers; many mistakes will come out because of the incorrect information from the eyes of workers. The all can lead wrong management of the wheel information. Therefore, it is very necessary to improve recognizing and examining methods of the wheels.The Machine Vision is an important method of non-contact measurements and developing rapidly recently. Using vision skill, information can be extracted, described and explained from the images of 3D environment. Combined with image processing, pattern recognition, and perceive science, the Machine Vision skill is widely used in automation, automobile, intelligent traffic, electronics and semiconductor.In order to solve the problems of recognition and detection of automobile wheels in domestic corporations, this paper gives a wheel recognition system that is used in the flexible produce line, based on the corresponding methods of image processing and pattern recognition. The recognition system can detect many wheels of different styles and different types, and the identification accuracy is greater than 99 percent.The main work in this paper includes:(1)Base on the real circumstance in the flexible produce line, the paper gives a method of high quality image capturing. After analyzing the reflection characters of different wheels shined with different color lights, appropriate light and background are selected, a vision system is also presented.(2)To the characters of different types of wheels, some feature extraction methods including histogram, edge, geometry and invariable moment based are considered.(3)According to the features, some theory and methods of recognition classify machines are discussed, and the nearest classifier and a voting classifier are designed and optimized in this paper.(4)Image processing and recognition software are finished and improved.The paper also proposes two criterions to measure the effective of picked features and gives accurate definitions;the concept of center histograms are brought forward and used to describe the structure of wheels;a kind of moment of inertia is proposed as an important feature which is sensitive to the size change but not to rotation and translation; at last, a voting classifier is given and used to the recognition and classify of various wheels.