Intelligent Visual Inspection Technology For Flexible Leather Material Defects

With the development of society and the improvement of our quality of life,our demand for leather products is increasing.For example,leather shoes,wallets,car cushions and leather chairs commonly used in our daily life are all leather products,so we need to constantly improve the production quality and efficiency of leather products.However,in the process of leather processing and manufacturing,leather raw materials are often due to natural or man-made reasons,resulting in defects such as tabanus eye,shear injury,fracture,crack,hole,scar and so on.Direct processing of leather will reduce the quality of leather products.Therefore,it is necessary to detect the defects of leather raw materials before leather cutting,so as to avoid defects and improve the quality of leather products.At present,leather defect detection is usually marked by professional inspectors.This kind of operation mode has low detection efficiency and requires high professional level of testing personnel.Human factors have great inf luence on the detection results,which can not guarantee the accuracy of the detection results.Therefore,this paper introduces the method of computer vision recognition to carry out intelligent detection of leather defects,instead of manual work,which can not only improve the efficiency and accuracy of leather defect detection,but also save manpower and reduce production costs,so that the following leather typesetting and cutting processes can reasonably avoid defects and make leather processing more intelligent.In view of the low detection efficiency and accuracy of leather defect detection,resulting in the low production efficiency and product quality of leather products,this paper carried out the research on intelligent visual inspection technolo gy for defects of flexible leather materials.The main research work includes the following aspects :1.In depth understanding of the causes and types of leather defects,and then through the image of all kinds of leather defects to analyze the characteris tics of leather defects.From the appearance shape and obvious degree characteristics of leather defects,the leather defects are classified according to the image features,so as to better identify leather defects later.2.According to the analysis of the image characteristics of leather defects,aiming at the problems of various defect forms and large interference of leather texture background in leather defect detection,a convolution neural network algorithm for leather defect detection was proposed to detect leather defects and complete the accurate classification and location of defects.According to the characteristics of leather defects,the algorithm model designed a suitable feature extraction network to extract leather defects better.The RPN network is used to efficiently select the leather defect areas,and then the softmax classifier and the regression based on NMS algorithm are used to identify,classify and locate the leather defects.Through the establishment of data set and model training,the accurate classification and location of leather defects are completed.3.The framework of leather defect detection system was constructed,and the visual inspection platform composed of light source,linear CCD camera,lens and actuator was built,and the visual inspection system was calibrated.Then the software framework of leather defect detection is designed,including leather defect detection module and camera control module.According to the trained leather defect detection model,the leather defect detection experiment and test are carried out.