画像処理を用いた経編ニット製品の糸切れ欠陥検出の自動化

Abstract

The factories to manufacture the cloth textile products need an automatic defect detection system to improve efficiency. However, the warp knitted fabric often used for sports etc. uniforms has a complicated structure. Therefore, the defects of warp knitted fabric cannot be detected by the existing methods such as the methods using a laser beam. We had proposed a method to detect the defects of warp knitted fabric by using block division and two different filters. This method needs a number of parameters, such as coordinates of perspective transformation, the width of block division, and a revision value of the threshold value at the time of judgment. Therefore, it has the problem that a number of parameters needs to be set manually. This paper proposed an inspection method using the block division and two kinds of filters which does not need the manual setting before inspection. The procedure is as follows: 1. The angle of the texture pattern is calculated from the image of cloth captured by a camera. This captured image is rotated based on this calculated angle. 2. The suitable block width is determined by calculating the cycle of a cloth pattern of transformed image. 3. The existence of a defect is detected without setting the threshold value manually by comparing the tendency of distribution of the filters output with a normal distribution. Based on the experimental results, it was confirmed that the defects of different types of cloth were correctly detectable from the image captured by a camera attached to a weaving machine placed in factory by using the proposal method.