Circle Detection Algorithm And Parametric Data Management

Circle detection is one of fundamental issues in machine vision, image processing andpattern recognition. Circular shape is one of the most common geometries in real life andindurstrial application. Reserch on recognizing and detecting circular objects in image can makesignificnt effect on pratical domain and has a great of application prospects in the domain ofindurstrial manufacture. Thus, a circle detection algorithm with less computation, higheraccuracy and beter stability is not only worth investigating, but also worthy of practicing.This thesis focused on circle detection algorithm in terms of speed and accuracy. Houghtransform, randomized Hough transform and randomized circle detection were deeply concerned,and several explorations and modifications were addressed on the basis of experiments. Theresults of paper could be good for the detection of circular hole of pin. Afew achievements are asfollows:1. Put forward a bias problem that exists in two dimension circle Hough transform.Accelerated randomized Hough transform is proposed for speeding up circle detection, anestimated formula that well explains the principle hid in the proposed algorithm is derived byusing generalized Bernoulli process. Statistical data from experiments demonstrate the exactnessand effectiveness of the proposed method.2. A modified randomized Hough transform is used to detect circular hole of pin, andseveral experiments are designed on the request of accuracy and repeatability precision, thealgorithm would be over on the condition that the sharpness of accumulation and the changingprobability tend to stable. Experimental results reveal the value of the proposed algorithm.3. Based on the analysis of randomized Hough transform and randomized circle detection, anew circle detection algorithm which is in the framework of recognition-refinement is proposed.Results form application demonstrate that this framework manages to balance well the trade offbetween speed and accuracy, and show the efficient of the proposed algorithm...