Research On Defect Detection System Of Molding Products Based On Monocular Vision

Molding product is now playing an important role in our daily life,and the requirements for its surface defect detection are getting higher and higher.Manual detection is the most common way to achieve this goal,however it's not very stable for the industrial standard.Compared with manual detection,detection based on machine vision has more advantages,such as fast,stable,non-contact and wild range of application.Therefore,machine vision has been paid more attention in the field of defect detection for molding products.Though there isn't a good solution to this problem due to the irregular shape of the molding products and the high reflection on the surface.In this paper,molding products are taken as the research objects,and the accurate detection of its surface defects is achieved by design specific illumination system,motion system,monocular vision system,and detection algorithm.By analyzing the types and characteristics of surface defects,two different visual detection platform are designed for different molding products,thus high reflection that is easy to occur on the surface of products can be eliminated.The uneven illumination in the images captured by the platform has been analyzed and corrected,and then the surface defects are segmented and marked,Thereby defect detection is realized.At last 3D reconstruction of the molding products is done,and then the defect is located in three dimensions.The main job of this paper is:1.The common surface defects and the requirements for them are analyzed firstly,and two different visual detection platforms are designed for different molding products.For cylindrical products,a rotary-type detection platform for high reflection products is developed.The whole side surface of cylindrical products can be detected,and the high reflection in the surface can be eliminated by this platform.For plate products,a flipping-type automatic detection platform is designed,it realizes the whole process including feeding,turning products over,detecting defects and sorting.It can also be docked with the pipeline to realize the whole automatic process from production to defect detection of molding products.2.Images of molding products are captured by the detection platform,and then preprocessing is done for the uneven illumination image after eliminating high reflection.An uneven illumination image enhancement algorithm based on correction co-efficient of row pixel intensity is proposed to eliminate the uneven illumination phenomenon.Next,the common algorithms of defect edge detection and defect image segmentation are analyzed.At the same time,a threshold image segmentation algorithm based on two times location is proposed.This algorithm can detect and locate defects accurately,such as whitening and speckle,and its application is verified by experiments.3.Considering the importance of 3D information for defect detection of molding products,the 3D reconstruction method for molding products is studied.Firstly,different feature extraction algorithms are introduced,and SIFT algorithm is used to extract features in this paper.A feature matching algorithm based on dual constraints is proposed.Then the sparse point clouds are reconstructed by SFM.Finally,dense point clouds are reconstructed by CMVS and PMVS.The high reflection on the surface of molding products are eliminated by means of hardware,the uneven illumination in the images are solved,the accurate detection of the surface defects of molding products are realized in this paper.Finally,the product is reconstructed,and the preliminary location of the defects is realized.