| Corner detection is a basic problem in the field of image processing, which is an important method of low level image processing.Corners are important image features that may be used at subsequent levels of computer vision such as object recognition,stereo matching and motion estimation.This paper provides an ALFM corner detection algorithm which is estimated with numbers of experiments and corner criteria. Compared with Harris corner detection algorithm,ALFM corner detection algorithm gains better results and stronger robusticity.Modality analysis and description are important methods in computer vision.And in the field of image processing,skeleton which includes important information of the object modality is broadly used in the fields of modality analysis and pattern recognition.The shape object recognition method based on skeleton is an important aspect of detection methods based on shape.Skeleton which is a simple and intuitionistic object description method uses the information of inner and outer region and roundly describes the planer essence of object shape.Compared with the skeleton,because approximate skeleton contains fewer nodes and arcs and its nodes and arcs are made up of lines,the structure of approximate skeleton is simpler and faster to query the edge length,which made the real-time fiber recognition system feasible.In the abnormity fiber recognition system,we need to calculate the approximate skeleton according to the corner and deal with the fiber recognition based on the approximate skeleton.Two algorithms,the ALFM corner detection algorithm and the abnormity fiber recognition algorithm based on SS,are provided in this paper.The results of experiment suggest that their recognition rates are similar to normal fiber,but the recognition rate of the algorithm proposed by this paper is better to gradient fiber than the SVM recognition algorithm.As a result,the algorithm proposed in this paper is more appropriate to the real-time fiber recognition system. |
