Research Of Fabric Surface Defect Detection System Based On DSP And FPGA

The automatic defect detection has been a research focus in the field of fabric quality detection on-line. The manual detection method has the low efficiency and the high false detection rate and the high rate of undetected defect. In recent years, digital signal processing technology and image processing technology are developing rapidly, which greatly boostes textile equipment’s intelligent and informational research. Based on the actual needs of the textile industry, this paper focuses on the key techniques and algorithm required by the fabric surface defect detection system.In order to overcome this problem that the fabric detect detection has the low detection accuracy and is difficult to achieve real-time detection, this paper constructs a detection system with a main processor DSP and a coprocessor FPGA. This system which is consisted of image capture, FPGA, image storage, image processing, data storage, peripheral circuit five modules can take full advantage of the digital signal processors’ high-speed operation ability and improves the speed of detection effectively.Fabric surface defect detection algorithm is the key of textile quality control. For textile texture features, this paper uses four methods which are adaptive thresholding segmentation method, iterative method, 2-mode method, otsu method to detect. Iteration method can separate the main part of the target and background, but it is not obvious to process some details of the image; 2-mode method can only segment the image whose histogram has two peaks; OTSU method can’t detect the complex fabric surface texture well. This paper proposes an algorithm based on adaptive threshold segmentation for defect detection. This algorithm enhances the detection probability of the threshold value corresponded to the fabric images’ gray histogram troughs and improves the accuracy of detect images’ segmentation. The defect detection rate reaches 93.6% through cambrayon detection experiment using the algorithm based on the hardware system platform. The result shows that the system can achieve the automatic and real-time fabric defect detection and has the high detection rate.This paper introduces a dedicated image processor to textile industry, to propose a solution which can achieve real-time detection in textile industrial site. The detection system can be applied widely and has the good scalability. It can be developed continuously and applied in other textile quality testing machinery, to achieve the real-time fabric quality detection.