Curve Identification Method Based On Modern Optimization Algorithm

Curve identification is the basic task about the field in image matching and identification,image analysis and understanding and so on. It is widely use in the field of industry,kinesiology and biomedicine etc. The frequently-used methods of Curve identification such asHough transform, Rand Hough transform, based on Arc combination method and the analysisof cluster, etc.In the paper, the issue of curve identification about multi-circle or arc will be solved withthe method of cluster. First of all, the analysis point at the drawback about the largecomputing capacity and the slow convergence rate in the traditional GA-FCSS, and improvethe select initial population in the GA step. The article is propose a new method to selected the initialpopulation which is using the method of “three points define a circle”. The experiments show that thenew method has better effects in the computing capacity and the convergence rate than the traditionalGA-FCSS in the cluster analysis issues of concentric circle.As a result of the FCSS is a method of curve identification based on FCM, it succeedsthe drawback of FCM such as sensitive about the initial value and nose point, the resultingmembership can’t represent the typical of classification in the data point, etc. For this reason,the article analysis the cluster algorithm FCM-PCM, and introduce the algebraic distance intothe algorithm. There propose a circle or arc identification method IPCSS based on the modelof FCM-PCM. The simulate experiments show that the IPCSS algorithm has a better clusterresult in the data point of concentric circle type, and it can solve the issues of sensitive about theinitial value and nose point to a certain extent.