Research On Leather Defect Detection Method Based On Infrared Thermal Imaging Technology

Leather-made shoes,bags,clothing and other commodities have penetrated people's lives and become an important economic pillar of leather production.Therefore,leather quality inspection is a very critical step in the leather production process of an enterprise.At present,many leather factories have introduced quality inspection technology based on machine vision,but the current inspection methods can only detect surface defects(scratches,cracks,penetration holes,etc.)that are visible to the naked eye.When the front of a piece of leather contains surface defects,the back When the defects include non-penetrating holes,it is impossible to detect all the defects at one time,which makes the detection efficiency low and the detection defect types are not comprehensive.The key research content of this paper is the research of leather defect detection algorithm.Aiming at the lack of existing detection algorithms and the characteristics of leather defects,this paper proposes a leather defect detection method based on infrared thermal imaging technology.The method first collects the visible light image and infrared thermal imaging image of the leather,and then uses the fusion algorithm to fuse the images.The fused image obtained after the fusion contains holes and defects on different sides of the same leather.Through holes and non-penetrating holes)and surface defects(scratches,cracks,spots)information,and finally through the defect detection algorithm,the detection results can be output at one time,shortening the detection time.The main research contents and contributions of this paper are as follows:(1)The leather defect detection system was designed,the overall structure diagram was drawn,and the control system was designed according to the system workflow and the functions of each mechanism.The FLIR A320 infrared thermal instrument and the VCC-G20E20 industrial camera are selected to form the image acquisition system,and the halogen lamp is used as the thermal excitation source.When designing the control system,the Mitsubishi FX2N-80MR-D is connected with the computer serial port as the controller,the input and output addresses are assigned according to the system function,and the program is written using GX Developer software.(2)A data set of common leather defects was established through industrial cameras and thermal imaging cameras.The leather defects detected in this paper are divided into five categories:penetrating holes,non-penetrating holes,scratches,cracks,and spots.An image acquisition system was used to collect leather images,and the size of the images was uniformly made to 320×240 for post-processing.In this paper,when collecting infrared thermal imaging images of leather,a comparative experiment is set up.Through the analysis of the experimental results,it is determined that the 800 w halogen lamp is used as the excitation source,the heating time is 8s,and the 12 th,13th and 14 th are collected.The frame image is used as a defect detection image.(3)A leather image fusion algorithm based on the Laplacian pyramid is adopted,and the fusion rules are formulated.Through the analysis of the collected leather defect data set,it is found that for the same leather that contains multiple types of defects at the same time,the infrared thermal imaging image only collects the defect information of penetration holes and non-penetration holes,and cannot collect other defect information.Visible light images can collect information on all types of defects except for holes that are not penetrated on the back.In order to be able to detect all types of defects at once,this paper adopts an image fusion algorithm based on the Laplacian pyramid.Through this algorithm,the infrared thermal imaging image and the visible light image of the leather are merged into one image,which makes the defect information more abundant and comprehensive.First,the two images to be fused are decomposed into a base layer and a detail layer with different resolutions,then PCA(Principal Component Analysis)and weighted average fusion algorithms are used to fuse the corresponding layers,and finally the fusion containing the information of the two images is obtained through pyramid reconstruction.Image.The sub-images obtained after the pyramid decomposition can completely retain the image structure,so that the sub-images of different layers can reflect different features and detailed information.(4)Propose an improved AC salience detection algorithm.In order to improve the accuracy of defect detection in fusion images,this paper proposes an improved AC salience detection algorithm.The AC algorithm calculates the local contrast of a perceptual unit in different neighborhoods in the CIE Lab color space to achieve multi-scale salience calculation to generate a salience value,and then normalize the salience value to separate the defect information of the fused image from the background.Significant map.The algorithm only calculates the local salience value but not the global salience value,so that the generated salience map has a low degree of separation between the defect information and the background,and there are phenomena such as loss of defect information,incomplete details and edge information,and so on.Therefore,this article improves it,introduces global contrast,calculates the global salience value of the sensing unit in the entire image,and calculates the local salience value of the sensing unit in the neighborhood at the same time,by comparing the global salience value with the local salience value and normalizing it Process to generate the final salience map.The experimental results prove that the salience map generated by the improved algorithm has higher contrast with the background,the defect information is more complete,and the edge and detail information is clearer.The improved algorithm performs better,with a detection accuracy rate of 79%,a comprehensive evaluation index of 86%,and an average running time of 1.5463s/frame,which can effectively detect leather defects.