Shoes Based On The Generalized Hough Transform Detection And Vectorization Research

This paper is based on leather incising CAM system and made the shoe pats to be the main studying object. In this paper, we researched the vectorization of graphics and detecting technique. We also made the closing curves to be vector graph precisely and solved the matching problem of the shoe pats in the project.The research contents of the paper include:1) Proposing a new algorithm of vectorization. Because the cutter can only cut vector graphs and to the same shoe pat it is not convenient to change the size using the bit map. So we should do the before cutting.This paper researched the theory and method of graphics' vectorization, analyzed the merits and defects of different algorithms and proposed a new curve-fitting algorithm based on key points. It searches the key points between tow end points using iterative method. If tow new sequent key points are added, the former key point between them will be deleted to reduce the number of the key points. This approach could control the precision by some threshold.2) Using the GHT (General Hough Transform) algorithm to solve the discontinuity and distortion problems of the shoe pats. We made normal templates to replace the anomalous shoe pats. The second problem belongs to edge detection problems. The traditional methods are based on space operation. But these detecting operators which are based on space area are very sensitive to noise and often enhance the noise when detecting. Because the GHT algorithm is a detecting method which is based on the images' whole characters, it is not affected by these factors. In order to solve the discontinuity and distortion we proposed a method that based on GHT to detect these anomalous shoe pats. When these anomalous shoe pats are detected, they will be replaced by normal shoe pats. So that normal shoe pats will be cut by the cutter. By analyzing the GHT' character and the task' requirement, this paper proposed an improved GHT algorithm which uses the normal direction as its index. This method can avoid the unsuccessful detecting which is induced by discontinuity and distortion. According to the testing result, this algorithm is simple, precise and can meet the requirement of the industry visual processing.