Research On Pose Measurement Of Stacking Workpieces Based On Machine Vision

With the continuous improvement of the level of automation and intelligence of manufacturing,the application scope of machine vision technology is more and more extensive,especially in the fields of automatic handling,division and recognition of workpieces,detection and measurement of workpieces in industrial production,visual technologies such as digital image processing,visual measurement,and visual recognition has become important components.In this paper,the stacking workpiece in industrial production is studied.The main purpose is to realize the recognition of the target workpiece and the measurement of its position and posture in stacking workpieces,which lays a foundation for subsequent work of grasping and transporting.Firstly,through the discussion of camera imaging theory and binocular stereo vision measurement principle,combined with the actual research background,the pinhole camera imaging model and parallel binocular stereo vision model are constructed.With the calibration checkerboard images collected in the experimental environment,and using the Zhang Zhengyou calibration method in MATLAB software to calibrate the left and right cameras,obtain the relevant internal and external parameters of the measurement system,correct the distortion of the collected images,and the epipolar correction for the left and right cameras images is completed.Secondly,according to the stacking situation of workpieces,the image pre-processing related technique is used to eliminate the image noise and highlight the edge feature information of the stacking workpiece.After comparing the results of stacked workpieces detected by several edge detection operators,an improved Canny edge detection algorithm based on interval two-type fuzzy sets is proposed,through actual experiment of detection,it shows that the algorithm can detect the edges between workpieces very well.Then,combining Hough transform circle detection method with random Hough transform ellipse fitting method,the edge features are roughly detected first,and then the ellipse is accurately fitted by dividing regions.The results show that the fitting effect is good.Then,stacking workpieces are classified into bare graspable workpieces and occluded non-graspable workpieces by using SVM classifier.A method of combining HOG features and noise-resistant multi-scale LBP features is proposed to train the classifier of support vector machine.The classification accuracy of the classifier is higher than that of the classifier based on single feature.Then the edge features of the identified grabbable workpiece are reconstructed,and the spatial coordinates of the center of the top circle of the workpiece and the angle between the workpiece and the vertical direction are calculated by fitting the spatial circle.Finally,the hardware system of stacking workpiece pose measurement is constructed,and a software system of measurement is developed and implemented.The system measurement accuracy verification experiment and pose measurement experiment are carried out.The experimental results show that the algorithm is reasonable and feasible,the measurement accuracy is high,and it can meet the needs of engineering and achieve the ultimate purpose of this paper.