Research On Intelligent And Automatic Inspection Machine

Fabric defect detection is a very important process in the production of the textile, which is the key to ensuring product quality and improving the level of quality control. Traditional defect detection mainly relies on artificial, which have many disadvantages, such as high missing rate, low productivity, do not meet modern people-centered concept of the production and so on. It is urgent to carry out information technology and automation in the entire fabric defect detection industry. To this end, the research group which author studies in developed a set of Machine Vision-based Platform for Real-time Fabric Defect Detection independently, which provided a platform for verifying the real-time of defect detection algorithms. The main work of this paper is researching on the now existing fabric defect detection algorithms, and finding one or several kinds of algorithms which have good real-time performance and good comprehensive detection abilities based on the automatic fabric inspection platform.First, the paper analyzes the background and significance of the issue, and makes a simple introduction about the automatic fabric inspection platform.Secondly, the paper verifies the feasibility of the now existing nine kinds of fabric defect detection algorithms, and develops nine sets of software in Visual Studio2008with ANSI C to realize all nine kinds of algorithms. Then the paper carries out the detection experiment for the same batch of plain fabric defect images with nine kinds of detection algorithms in offline condition. After that the paper selects the superior algorithms from the detection results, the result: local threshold segmentation algorithm, least-squares fitting algorithm, two-dimensional fractal algorithm and one-dimensional fractal algorithm are four kinds of defect detection algorithms which have good comprehensive detection abilities.Thirdly, the four superior algorithms are embedded in DSP, after that, each algorithm are applied to online detection for the same plain fabric, the result are analyzed and summarized carefully. Finally, the real-time of the four defect detection algorithms:two-dimensional fractal algorithm> one-dimensional fractal algorithm> least-squares fitting algorithm> local threshold segmentation algorithm.Fourthly, the paper makes a carefully analysis about the shortcomings of the traditional defect detection algorithms which based on the detection template, and then puts forward two kinds of improved template based defect detection algorithms:defect detection algorithm based on grayscale statistics and morphological opening operation, defect detection algorithm based on grayscale statistics and connected domain area statistics. The paper develops two sets of software in Visual Studio2008with ANSI C to realize the two kinds of algorithms, and carries out the detection experiment for the same batch of plain fabric defect images with two kinds of detection algorithms in the offline condition, the detection results are analyzed and compared carefully, and then concludes:both algorithms can be successfully used for fabric defect detection, the overall detection performance of the former is superior to the latter.Finally, the paper makes a comprehensive analysis and comparison about the detection result of the two kinds of algorithms which proposed by author and the four kinds of algorithms which optimized from the existing algorithms. Finally result:the defect detection rate of template based defect detection algorithms is higher than algorithms with non-template, and the overall detection performance of the former is superior to the latter, template based defect detection algorithms is the main development direction in future. In addition, effective optimization for algorithms has a significant impact on the improving of the defect detection efficiency.