Automatic detection and precise location of texture image defect based on ambiguity resolution algorithm

Abstract

Most of the detail structure characteristic information of the texture image is lost due to the space variation during the automatic detection, which reduces the precision of auto detection and precise location. In order to solve the problem, a method based on the ambiguity resolution algorithm is proposed in this article. The background texture information was eliminated via the Gaussian filter and threshold value method to acquire initial rough edge of the defect area. In the stage of precision location, the modified Chan-Vcsc active contour model based on the fuzzy energy was introduced and the precision location for the defect area was achieved via level set method using the rough edge as the initial curve of the model evolution. Experimental results show that the proposed method can achieve accurate and automatic localization of different types of defect areas, and has high computational efficiency. When the number of samples is 600, the positioning time is only 19 s. When the peak signal-to-noise ratio reaches 100 dB, the absolute error of this method is only 0.01%, the relative error is 0.04%, and the positioning accuracy is as high as 99.58%.