基于方法库的织物图像疵点检测

Fabric Image Defect Detection Based on Method Library

为准确检测织物在生产过程产生的疵点,提出一种基于改进的Gabor滤波方法、数学形态学处理法和多尺度小波检测的方法库的系统检测法.首先采用改进的Gabor滤波方法,选出最优滤波结果,进行高斯平滑,确定正常织物图像的两个阈值门限,进而分割出织物的疵点图像;其次采用数学形态学处理法对织物图像进行检测;最后采用多尺度小波检测的方法,检测最终结果.由于织物的纹理不同,在生产过程中产生疵点的种类众多,算法采用级联检测,保证了检测疵点的准确有效性.试验证明,所提出的算法检测结果较好,能准确定位疵点的位置.

To detect the fabric defects in production process accurately, a method library defect detection approach which was based on the improved Gabor filter method, morphological operations and multi- scale wavelet detection was proposed. Firstly, the optimal filtering results were selected using the improved Gabor filter method, and the two thresholding of the normal fabric image were determined through the Gauss smoothing, then the defects were segmented. Secondly, the morphological operations were applied to detect defect. Finally, multi-scale wavelet detection method was used and detected the final detection results. Numerous defect species are produced in the process of production due to the different fabric texture, so the cascade detection algorithm was applied to ensure the effectiveness of the defect detection accuracy. The experiments show that the proposed algorithm detection result is better and the defects are located accurately.