关键词: 铁路扣件, 边缘特征, 高斯拉普拉斯算子, 布谷鸟搜索, 莱维飞行

Abstract: Edge character is the important basis of rail fastener defect classification. The edge extracted by LoG operator is most close to the real edge of fasteners, but the effect of edge detection is directly affected by the selection of LoG parameters. In view of the LoG parameter optimization, optimizing the parameter of LoG algorithm based on the cuckoo search optimization is adopted. First of all, adjacent pixel threshold value and standard deviation parameters are optimized based on the cuckoo optimize search operator to improve the edge detection performance of LoG operator. Then, the quality factor of Pratt IMP value is used to judge whether the edge character is the best one. Finally, the experiments show that this method is effective to optimize the parameters of the LoG settings. The accuracy of fastener defect classification is improved based on the improved edge character.

Key words: railway fasteners, edge?character, Laplace of Gaussian（LoG）, cuckoo search, Levy flight