Research On Application Of Machine Vision Technology In Precision Detection

Floater module is a key component of gyroscope,and its static unbalance directly determines the accuracy level of gyroscope.The static unbalance of the floater can be obtained by measuring the radial rotation angle and the axial tilt angle of the floater in the floating fluid.The current detection method is manual detection,that is,the movement of the floater in the floating fluid can be observed by experienced operators with the naked eye.In this detection mode,human subjectivity has a great impact on the detection accuracy,and the repeatability of detection is poor.A machine vision precise detection system for floater attitude angle which is suitable for practical engineering application is designed.The selection basis of hardware modules involved in the system(such as image sensor,lens,light source,etc.)is discussed in detail.The hardware selection,hardware matching and structure construction scheme of the system are proposed.The design and development scheme of software modules of the system are also given.A region of interest(ROI)extraction algorithm for floater radial end surface image is studied,and a ROIextraction algorithm with high robustness,scalability and efficiency is proposed.The algorithm finds out the concentric circle texture features of the floater end surface by border following,ellipse fitting and DBSCAN clustering,and combines the prior informationto calculate the inner and outer elliptical boundaries of the region of interest.The closed area surrounded by the inner and outer boundaries is segmented to obtain the ROI of the radial end surface image of the floater.Based on the results of ROI extraction,two kinds of floater radial rotation angle detection algorithms,which are based on Hough transform and SIFT,are designed for different stages of static balance.The main difference between the two algorithms is that the detection and matching algorithms of feature points are different.The first algorithm uses Hough transform to find the circular texture features in the image and assigns each circular texture feature a unique marker.The feature matching is performed according to the marker correspondence in the two images.The detection accuracy of the algorithm is less than 0.6 degrees,and the repeatability accuracy is less than 0.2 degrees.The second algorithm uses SIFT to find feature points with scale invariance and rotation invariance in the image.The Euclidean distance between the feature point descriptor vectors is used for feature point matching,and the matching results are filtered by RANSAC and other methods.The detection accuracy of the algorithmis less than 0.02 degrees,and repeatability accuracy is less than 0.006 degrees.An algorithm for detecting the axial tilt angle of a floater is proposed.The algorithm divides and extracts the side profile of floater by border following and distance transformation.The approximate profile of the floater side profile is fitted by the minimum outer rectangle.The minimum outer rectangle angle is used as the axial tilt angle of the floater.The detection accuracyof the algorithm is less than 0.5 degrees,and the repeatability accuracy is less than 0.02 degrees.The floater attitude angle machine vision precision detection system and the floater attitude angle detection algorithm proposed in this study have been applied in practical engineering,which can effectively avoid the subjective and individual differences caused by manual detection.It has the advantages of non-contact,visualization of results,good real-time performance,high automation,high detection accuracy and high repeatability accuracy,and has highengineering application value.