Real-Time Detection System For Automobile Oil Pump Supporting Bar Dimension Based On Machine Vision

The quality of automobile parts has an important impact on the overall performance of the automobile.It is necessary to evaluate the quality of the automobile parts after they are made up.Traditionally,the quality of parts is detected mainly by manual inspection or by the auxiliary machines.Because of their limitations,the detection accuracy is not high,and large-scale automated production cannot be achieved.So with this demand,the machine vision technology has been introduced into automotive parts' detection.The machine vision detection technology uses the industrial cameras to capture the images of the target object,and the image processing software to get the results of its detection.Then the related controller realizes the appropriate operation.This technology has the advantages of non-contact,real-time,high precision,and easy to automate the management features.Toward the automatic detection of automobile oil pump supporting bar,we propose the method to realize the detection by using the machine vision to replace the original traditional manual detection.In this thesis,the real-time detection system of the supporting bar is completed by designing the hardware and software architecture,image definition evaluation,sub-pixel-based support bar geometry measurement,and sorting control equipment.The software and hardware structure of the vision system adopts special light source module and the image sensor module.At the same time,a no-reference image definition evaluation algorithm NRFSIM(No-Reference Feature Similarity)is proposed.Because of the mechanical vibration of the real-time detection system and the relative motion of the supporting rods,there is the blurring in the subsequent captured images.In this thesis,The NRFSIM algorithm is proposed by applying a reference image quality assessment algorithm FSIM(Feature Similarity Index for Image Quality Assessment)to the no-reference image definition evaluation algorithm.The performance of the proposed algorithm is better than that of image definition evaluation algorithm based on the image gradient in the multi-scene image and the support bar image definition evaluation of the system.A novel algorithm is proposed for measuring the size of supporting bar based on the sub-pixel processing.Sub-pixel edge detection algorithm is based on the fitting method to detect the edge of the image.For the size measurement of the supporting bar,the algorithm of rectilinear range finding and point ranging algorithm are studied.Finally,the experimental results show that the point ranging algorithm of detection system is of higher accuracy.Nevertheless,the camera calibration method system is carried,and the real-time detection and sorting system for the supporting bar dimension is further developed,which has been applied in real-time production quality inspection.The results prove that the system has enough repeat precision and detection precision.The proposed algorithm and the design of the system can also be used for the inspection of other application areas of the industrial parts manufacturing.