关键词: 划痕检测, Gabor 滤波, 纹理抑制, 高斯函数, 各向异性, 滞后多阈值

Abstract:

Classic scratch detection usually uses variety edge operators. Because the edge detection alogorithms are sensitive to texture and noise,they often cause a lot of false positive results. In the case of the detection of metal surfaces, due to complex textures present in the surface of metal material,the false positive results are particularly serious. Here, based on bar pattern detection principle of Gabor filtering and combining with anisotropic texture suppression and hysteresis multi-threshold technology,a scratch detection method used for mobile phone accessories is proposed. First,the method extracts the scratches frame using Gabor filtering,and secondly,uses anisotropic texture suppression on the metal surfaces. Finally,it extracts scratches accurately with hysteresis multi-threshold technology. Experimental results show that the method can greatly suppress the texture of mental surface in background. At the same time,it extracts the complete scratch images. The false positive detection rate,false negative rate and probability of contour missing achieve 2% ,3． 7% and 5． 5% respectively,and the performance of the method is obviously superior to edge-based scratch detection methods.

Key words: scratch detection, Gabor filtering, texture suppression, Gaussian function, anisotropy, hysteresis multithreshold