Fabric Defects Image Recognition And Classification Based On (?)~0 Norm Optimization Algorithm

Defects diversity is an important criterion of the cord quality assessment. In the process of cord fabric manufacture we need to classify the detecting cord fabric defects. Due to the hardware conditions such as the precision of the sensor and the effectiveness of the detection algorithm, the traditional defect detection can only carry on the detection and recognition of fabric defects but ignores the category of the defects. To adopt new technology and intelligent algorithm, we can improve the accuracy and effectiveness of the cord fabric defect detection on a relatively lower hardware platform, which is of great significance for both scientific research and manufacture.The existing intelligent image recognition algorithm is most suitable for human face, fingerprint, and other aspects of medical images, and the development of these algorithms has been quite mature. Cord fabric defect image not only has the characteristics of ordinary image, at the same time also has its own particularity. So the relatively mature intelligent image recognition algorithm can be applied to the image recognition and classification of the cord fabric defects. The process of intelligent image recognition and classification based on machine vision is as follows: First you need to use the appropriate method to establish a cord fabric defect image library as a training set of image classification, secondly take recognition and classification of the images by recognition classifier. This article puts forward the weighted norm image reconstruction algorithm through learning the classical norm optimization algorithm. Increases the sparsity of the graphics by using the image sparsity at first and finally gets the classification information of the fabric defects image through designing a weighted norm classifier to classify defect image simulation.There are many existing cord fabric defect image recognition algorithm, such as Gray histogram, the BP neural network、threshold segmentation and so on. But each method has certain defects due to its own characteristics. For example: The error rate of the Gray histogram for image detection is higher, the calculation of the Bp neural network is so big. The proposed algorithm of defect image recognition and classification based on the norm optimization algorithm, to monitor and classifies the quality of the image by the use of the image sparsity, which reduces the data redundancy as well as the interference of outside noise, Thereby realize rapid detection and classification of defect under the complex background.