Based on the above reasons, the automatic detection of fabric defects is one of the hot topics concerned and studied by scholars at home and abroad in recent years. Using computer image processing technology can improve the accuracy, rapidity and comprehensiveness of textile detection, and provide guarantee for enterprises to control and improve textile quality online. Fabric defect image recognition refers to the method of processing the fabric surface image according to a certain algorithm to identify the type and degree of defects.
The research on automatic recognition of fabric defects abroad has been more than 20 years, and its development is relatively mature. The more influential systems are: Fabriscan system of Swiss Huste company, online fabric inspection system of German Obdix photoelectric technology company, and I-TEX system of Israeli EVS company. The composition and characteristics of these automatic cloth inspection systems are as follows:
(1) The Fabriscan automatic cloth inspection system of Huste Company adopts neural network technology. It takes about 1min in the initial learning stage to record the characteristic parameters of the normal appearance of the fabric at 1m, and then enter the inspection stage to find the local anomalies different from the normal appearance, and analyze, mark and record them. In addition, the test results can be input into the integrated quality management system to classify the defects and further evaluate the fabric quality.
(2) The online fabric inspection system developed by Germany Obdix Optoelectronic Technology Company combines optics and mechanics, and with the support of software processed by neural network method, the surface of the fabric being woven is inspected by sensors. The device can classify the following defects: dirty, holes, broken warp, broken weft, yarn jumping, knots, knots and color defects.
(3) I-TEX, the core equipment of the automatic inspection system for grey cloth of Israel EVS Company, is a set of observation inspection system. This system is based on an independent image understanding algorithm, which imitates the human visual mechanism and can automatically control the inspection, storage and positioning, and further evaluate and analyze the fabric defects. Visible defects as small as 0.5mm can be detected with a detection width of 330cm.