Image Segmentation And Image Recognition Algorithm For Cord Fabric Defects Inspection

This paper cuts up and recognize pretreatment of cord defects image segmentation . Image segmentation is a very important step in image processing , especially in image processing of detection system of the product online. There are many algorithms , but there is no algorithm of the image segmentation for all products. And the effect of image segmentation directly affects the feature extraction of the follow-up defects flaw and the identification of defects type. In this paper, for image segmentation and identification of defect types have adopted several improved algorithms for image processing , and select algorithms of the best effect and the best real-time based on experimental results .This paper analyzes and compares several of the most typical and common characteristics of traditional methods of the image segmentation and recognition of image classification, and select the optimal algorithms in each part together to constitute the image pre-processing segmentation, we used the classical Otsu method. We found that the simulation result of this method makes the defect distorted .Then threshold surface method based on the statistical theory was built. To address the threshold surface construction from a huge amount and accuracy of the calculation proposed, this method presents threshold surface construction method which is based on the direction of the statistics. At the same time, we also use the PCNN algorithm to cut up defects image. After the segmentation, we adopt a defects flaw clustering algorithm approach based on sequence and a long route in order to have a better understanding of the defects information. In the image recognition section, classical BP neural network is used to classify the type of defects while we compare with the improved classification method of SVM. Experiments show that the algorithm which we choose ultimately achieve better recognition results, and the relatively strong real-time.