Research On The Detection Technology Of Paint Surface Defects

Car paint is not only used as a protective layer for auto parts,its aesthetics degree often determines the user’s first impression at the time when people purchase.At present,the detection of automotive paint surface defects is mainly based on manual detection methods.This method has low efficiency,poor stability,and is greatly affected by human subjective factors,which seriously affects product quality.In recent years,the rapid development of machine vision has solved many problems such as difficult detecttion,false detecttion,and missed detecttion in surface quality detecttion.In view of the low detection efficiency,complexity system,poor stability and other problems of the current automotive paint surface detection,this paper takes the surface of the car hood as the research object,combines machine vision technology with mobile robot technology,and design and develop a complete set of applicable Large-scale paint surface defect detection systems,similar to automobile bodies.The vision system is driven by the manipulator to collect the complete image of the hood paint surface,and the defect detection is completed.(1)Complete the design of the software and hardware structure of the detection system.The technical indexes of the detection system are determined according to the different processing techniques of automobile paint,the common defect types of the automobile paint surface are analyzed,and the gray particles and scratches are selected as the main research objects.Combined with the surface imaging characteristics of the paint surface,the overall technology is designed.(2)Use visual guidance to complete the posture correction of the robotic arm.Zhang Zhengyou’s camera calibration algorithm and TASI two-step method are used to solve the handeye calibration matrix,the spatial conversion relationship between the end of the manipulator arm and the camera coordinate system is determined.According to the characteristics of the coaxial light source,combined with the results of the imaging of the painted surface,the offset of the center point and the rotation vector are calculated through the geometric characteristics of the image,and the automatic posture adjustment of the mechanical arm is realized.(3)Detect the quality of paint surface combined with pattern recognition algorithm.The image preprocessing operation is carried out on the collected paint surface image to enrich the characteristics of defects.An improved Local Binary Pattern(LBP)feature is proposed,which enhances the rotation invariance of the feature.According to the different characteristics of the defect,the defect feature parameters are selected,and 11 features including energy,entropy,contrast,inverse difference moment,projection histogram,and improved LBP histogram are established to form a 97-dimensional feature vector.Principal Component Analysis(PCA)is used to reduce the dimension of feature parameters and reduce complexity.The Libsvm software package is used to determine the optimal parameter pair,and the Support Vector Machine(SVM)algorithm is used to identify and classify defects.(4)Verify the performance of the detection system.Design and develop the humancomputer interaction software interface of the system,then designed experiments,and test the collected image samples.The experiment results show that the detection accuracy is 97.8%,the recognition effect of scratches is 93.33%,and the ash defects recognition effect is 94.67%,which meets the requirements of detection.