Machine Vision-Based Online Inspection Method for Weft Density of Woven Fabrics

Given the current situation that the weft density of woven fabrics mainly relies on manual detection and the accuracy of current image measurement methods is not high, this paper integrates the use of image processing, deep learning, and other methods for real-time detection of fabric weft density. Using deep learning, the defective images that have an impact on weft density detection are excluded, after which the normal images are pre-processed and the weft density contours are smoothed using mean filtering. Then the texture period of the image is calculated through the smoothed image, the corresponding filter kernel is designed according to the texture period, the image is convolved to obtain the feature map of weft density, and the number of corresponding weft roots in the standard unit length is calculated according to the feature map to complete the automatic output of the weft density results of woven fabrics and realize the automatic detection of weft density of woven fabrics. The test results show that the method can detect the weft density of woven fabrics accurately.