Research And Its Implementation Of Fabric Defect Detection Algorithm Based On GPU

As a traditional industry in China,textile industry is the basic part of national economy,With the continuous improvement of people's living standard,the requirement of clothing quality is increasing,so the defect detection of fabric becomes the key to measure the quality of clothing.The traditional method of fabric detecting is to observe the cloth with the naked eye through a high-density mirror,this method has long detection time and low efficiency.Moreover,the judgment of artificial vision is highly subjective and can reduce the detection rate of the defect.Based on computer vision,the method of automatic detection of defects in fabrics by image technology emerges,Therefore,the technology of fabric defect detection based on computer vision has attracted the attention of scholars all over the world in recent years.At present,most computer vision algorithms based on the frequency domain and spatial domain of fabric defect are implemented based on CPU serial programming,It can not meet the industrial strength and requirements of large-scale real time in the modern textile industry.In this paper,a parallel calculation method of fabric defect detection is proposed based on GPU,we use the parallel processing of multithreading features to compensate for the poor real-time performance of defect detection.This paper combined FFT,Hough transform,and neural network to detect the types of defects based on CUDA's GPU architecture,and analyzed The comparison of detection rate and detection efficiency with the detection result in serial modes.The results show that the fabric defect detection based on GPU can be used to detect fabric defects under three different algorithms,however,FFT and Houghtransform have some limitations on the type of fabric defect.Three kinds of fabric defect detection schemes based on GPU have greatly reduced the detection time and improved the efficiency of defect detection compared with the fabric defect detection in CPU serial mode.This topic provides new ideas and theoretical research methods for the online detection of fabric defects.