Research On Mobile Phone Surface Defect Detection Method Based On Machine Vision

Defect detection is an important development direction in the field of machine vision,among which surface defect detection is an important means to measure product quality.With the extensive application and development of defect detection technology,mobile phone surface defect detection is also in the stage of moving from manual detection to automatic detection.The problem of mobile phone defect detection based on 2D image and 3D point cloud has become a hot issue in the detection field.Compared with 3D defect detection technology,2D defect detection technology is more mature,and 3D defect detection technology has a very broad development prospect.In this paper,using the combination of theory and experiment,the first defect of two-dimensional image preprocessing and segmentation algorithm theory to study and improve,to make it more suitable for mobile phone screen defect detection,further research on 3D detection in defect detection,the application of the point cloud registration algorithm was improved,and improved algorithm feasibility is validated by experiments.The main research contents of this paper are as follows:(1)In order to reduce the influence of illumination on the phone's screen image acquisition,image grayscale and adopt bilateral filtering de-noising,adaptive Gamma correction based on Tetrolet transform algorithm for image enhancement,image by Tetrolet transformation into background and details,Gamma correction algorithm was modified using nonlinear weighting function to remove the background noise of light,Canny operator is used to analyse the defect edge to strengthen,enhance the visibility of the flaws and the background.(2)Aiming at defect detection of two-dimensional mobile phone images,a defect detection algorithm based on particle swarm optimization(PSO)combined with linear intercept Otsu segmentation is proposed.The inter-class variance function of the linear intercept is used as the adaptive function of the particle swarm optimization algorithm.By updating the speed and position of the example,the defects are iterated until the optimal threshold is output,and then the defects are segmenting.Defect features were extracted by finding the minimum outsourcing rectangle,and a decision tree based on feature statistics was established to detect and classify defects.(3)Optimization of 3D point cloud registration algorithm.SAC-IA registration algorithm was used to register point cloud images rough,and ICP algorithm based on slice normal vector was proposed to register point cloud images fine.The normal vector of the point cloud section is added to the error function,and the limiting condition of minimizing the error function is added to solve the problem of local optimal solution.In the VS2013 platform,OpenCV3.0 is used for preprocessing,defect segmentation,detection and classification experiments,which verify the superiority of the improved preprocessing algorithm and segmentation algorithm and the feasibility of detection and classification algorithm.By using PCL1.80 open source library and point cloud registration experiments,it is proved that the improved algorithm is better for solving the problem of local optimal solution,and the matching effect is better.