Research Of Steel Wire Heald Automatic Detection Based On Machine Vision Technology

Textile industry belongs to the traditional industry in our country; the domestictextile industry has entered a period of restructuring and upgrading. Since2012, thegrowth rate of domestic textile industry was slowing; one reason is the slowdevelopment of the equipment industry which associated with the textileindustry. Steel wire heald is a damageable part of small and mediumsized textile machine, it needs to be replaced regularly, and China has greater demandfor textile heald because of big textile industry. However, there is no qualitytesting equipment of wire heald in china at present, the method of artificial detectionis the only way for enterprises to adopt. Artificial detection is arbitrarily, easy fatigue,slow speed and easy leaked.In order to develop a steel wire heald detecting equipment, we proposed amethod with machine vision, this paper mainly include some work as following:(1)The present research situation of the steel wire heald quality detectionis introduced; detection process is designed based on the characteristics of steel wireheald. The detection process will be divided into conventional defect detection andcomplex defects. Only the first step is finished and the result shows that the productsis Qualified, the process can enter the second step. Otherwise mark detection result asnot qualified. Skip second steps, this method can reduce the system Consumption.(2)Built a machine vision detection platform; given a brief description of eachpart of the system; introduced the principle of the double-sided imaging device andthe Waste removing device.(3)The algorithms involved in the detection of the steel wire heald wasintroduced, especially introduced the algorithms of enhance and thresholdsegmentation.(4)In the process of detection, firstly, earrings width, cang angle and yabaiwhich were relatively easy defect have been detected. The detection algorithms werediscussed in detail. In the detection of cang angle, we proposed a method of rotaryspindle which has a better perform.(5)In the detection of complex defects, we propose a method of positioning theregion of hole (region of interest) by5geometric parameters. By locating the regionof interest, in order to narrowing the target range which is conducive to the follow-uptreatment; (6)For the detection of complex defects, we used the method of combine thegray level histogram, gradient feature and gray level co-occurrence matrix.Extracted56s features from the original image and the laws image respectively. In theclassification, used principal component analysis to reduce the dimensions of feature,Used the support vector machine as classifier. Compared the classification resultsof linear kernel function, Sigmoid kernel, polynomial kernel function, radialproduct kernel function. The test results show that Kernel function with radial is betterthan others, the method of combination of two types of features is better than usingthe method of single feature.(7) Realized the detection algorithm and software, completed the software andhardware testing, and completed the preliminary debugging in the field of industrial.