Design And Implementation Of Machine-vision Based High Precision Detection System For Piston Shape And Position

Piston is one of the essential basic parts in the automobile engine,its quality plays a decisive role in the stability and performance of the engine.Therefore,the piston must undergo strict inspection of its shape,size and defects before leaving the factory.At this stage,the detection method of piston shape and position is still mainly contact detection,which is difficult to detect all the pistons that are shipped and has low efficiency.In contrast,the piston quality detection based on machine vision can detect multiple quality indicators at the same time.This method has the advantages of high efficiency,high precision and non-contact and realize the automatic detection of the piston quality.In this paper,a high-precision shape and position measurement system of piston is developed.The shape and position measurement of pistons’ key parts is realized based on multiple image processing subsystems.For diameter measurement of top circle,bottom circle and pin hole,by changing the field of vision,it solves the problem of large size high-precision measurement and completes the corresponding roundness detection;for the inner edge distance of pin seat,by multi-scale threshold segmentation,it solves the problem of difficult to locate the target edge and poor noise resistance;for the coaxial detection of pin holes,by combining the advantages of deep learning and traditional detection,it solves the problem of single feature extraction and poor adaptability.The main work is as follows:1.Aiming at the multi class dimension detection of piston,two camera calibration dimension measurement method and multi-scale distance measurement method of pin seat inner edge are proposed.For the top surface circle,bottom surface circle and pin hole diameter,which belong to large-scale and high-precision measurement,we convert it to small field of vision precision measurement by dual camera calibration dimension measurement method.That is to say,when the resolution of the camera remains unchanged,the field of view is reduced to improve the single pixel accuracy.In order to improve the accuracy and efficiency of edge location,a coarse to fine edge location method is introduced in the measurement;for the inner edge distance of pin seat,which belongs to the part of small size and high precision measurement,we use multi-scale threshold segmentation to accurately locate the edge position,and then realize the distance measurement.2.Aiming at piston shape and position detection,a pin hole coaxial detection method based on dynamic area analysis is proposed.The roundness of top circle,bottom circle and pin hole can be measured in the corresponding diameter measurement.For the coaxial detection of pin holes,it is difficult to realize directly by machine vision,so we transform it into a dynamic analysis of the moment when the pin hole area reaches the maximum value.That is to say,let the piston rotate,and then use two cameras to collect the images of the pin holes on both sides,and judge whether they are coaxial by calculating the time difference when the area of the pin holes on both sides reaches the maximum.The key to calculate the area of pin hole is to locate the location accurately.Therefore,a pin hole location method from coarse to fine is designed by combining depth learning with traditional algorithm.3.On the basis of the above two works,the design and development of the piston highprecision shape,position and size detection system are completed.In hardware,two detection stations are built,and appropriate cameras and light sources are selected.And in software,the system architecture,operation interface and background management are completed,and the detection algorithm is integrated into the system.The automatic non-contact detection of piston shape,position and dimension index is realized,and the detection efficiency is effectively improved.