Research On Detecting Small Modulus Worm Surface Defect Based On Machine Vision

Worm is an important machine part in industrial production.Affected by processing technology and production environment,it is inevitable to produce some surface defects on worms.If the defective products are not eliminated,not only the mechanical properties and life span of the products will be seriously reduced,but also the safety risks may arise.In industrial production,small modulus worm output is huge,detection efficiency is very important for worm quality detection.However,artificial visual testing is still widely used in enterprises.The defect detection efficiency is low and the reliability is poor.Machine vision detection technology is more and more widely used in the detection of surface defects of industrial products because of its advantages of non-contact,high speed and high stability.In this paper,the detection technology of small modulus worm surface defect based on machine vision is studied.The main works are as follows:(1)A worm surface image acquisition system based on linear array camera is built.According to the characteristics of worm surface defect,the whole layout of worm surface defect acquisition system is designed.The linear image sensor is used to scan and collect the worm image.A worm image acquisition system is set up,and the hardware parameters are set up for experiments to determined the scanning position of the linear camera.(2)Worm image preprocessing algorithm and surface defect extraction algorithm are studied.The adaptive median filtering method is used to deal with gauss noise and pulse noise.According to the characteristics of image gray value,the threshold segmentation method of binary threshold is used to segment the image.For different types of defects,two morphological processes are designed to extract the shape and feature of defects.(3)The algorithm of worm defect classification and recognition is studied.The geometric features and histogram features of the image region are extracted.Random forests are used to sort the importance of feature extraction and to select the features that contribute a lot to the classification.Support vector machine(SVM)is used to identify and classify defects.Train the parameters of SVM model with improved particle swarm optimization(PSO).(4)The performance of worm surface defect detection algorithm is studied.The defect feature data is normalized to avoid the influence of the order of magnitude on the classification results.In order to verify the performance of the defect detection algorithm,the classification accuracy of the algorithm is tested,the result is verified by the 5-fold cross validation method,and the time efficiency of the worm surface defect detection algorithm is tested.The average time consuming of filter noise reduction,image segmentation,defect extraction,feature extraction and classification in the process of defect detection is counted.The influence factors of brightness of light source brightness,rotational speed and vibration are analyzed.