基于Gabor滤波的重建图像法检测皮革表面缺陷

Gabor filtering-based reconstruction image method for inspection of leather surface defects

针对经典的缺陷检测算法无法解决纹理复杂、对比度低的皮革表面缺陷检测的问题,提出一种基于Gabor滤波的重建图像算法。首先设计Gabor滤波器,从纹理图像与特定的Gabor滤波器的卷积中计算能量的输出响应,当纹理特征对应的能量最大时确定Gabor滤波器的最佳参数。再将原图和滤波后的背景图相减重建图像,重建图中纹理和缺陷的灰度值差异被增强,通过中值滤波除去噪声和部分纹理,最后经过图像分割后计算灰度共生矩阵的能量,缺陷能量较低而纹理等干扰的能量较高,设定简单的能量阈值就可以实现缺陷区域的识别和定位。对大量皮革图像进行测试,实验结果表明:该方法可以准确、高效地检测皮革表面缺陷。

Aiming at the problem that typical defect detection algorithm cannot solve the defect detection of leather surface with complex texture and low contrast,a Gabor filter-based reconstruction image algorithm is proposed.Firstly,the Gabor filter is designed,and the output response of energy is calculated from the convolution of the texture image and the specific Gabor filter.When the energy corresponding to the texture feature is the largest,the optimal parameter of the Gabor filter is determined.Then subtract the original image and the filtered background image to reconstruct the image.The gray value difference between the texture and the defect in the reconstructed image is enhanced.The median filter removes the noise and part of the texture.Finally,after the image segmentation,the gray level co-occurrence matrix has calculated The energy,the defect energy is low and the texture and other interference energy is high,and setting a simple energy threshold can realize the identification and positioning of the defect area.A large number of leather images are tested,and the experimental results show that this method can accurately and efficiently detect leather surface defects.