Research On Wheel Hub Type Automatic Identification Technology Based On Depth Feature Sequence

With the concept of carbon neutral proposed,the automotive industry ushered in structural adjustment,but also reaped new market demand.As one of the key parts of automobile,the production of wheel hub is gradually moving towards automation.In the mixed hub production line,the scheme of manual identification of hub model has been difficult to meet the needs of industrial automation,so the automatic identification system of hub model needs to be developed to improve the production efficiency.Aiming at the problems existing in the manual recognition and traditional image recognition scheme,this paper carries out the research of automatic recognition of wheel hub model by using three-dimensional point cloud.Based on analysis of system requirements,the technical route of data acquisition,feature extraction and template matching is designed,and the overall structure and workflow of the wheel type identification system are determined.The 3D point cloud acquisition system is constructed to realize the detection of wheel hub position and the automatic acquisition of 3D point cloud.The massive invalid data points in the collected point cloud are divided into two different kinds.For the interference point cloud,an adaptive filtering scheme based on the distribution of point cloud amount is proposed.For outliers,the filtering effects of radius filtering and statistical filtering are compared and analyzed,and the effective removal of invalid data points is achieved.Considering the shape features of hub and the characteristics of point cloud data,a depth feature sequence extraction scheme is proposed as the basis of hub type identification,and the Pearson coefficient is introduced to verify the correctness of the feature extraction scheme.In order to extract the depth feature sequence,it is necessary to find the center position of the hub.In this paper,a scheme is designed to fit the center of the hub by projecting the point cloud onto the two-dimensional plane,and compared with the scheme of directly fitting the center of the hub from the point cloud using RANSAC circle algorithm.After feature extraction,the KNN classification algorithm with checking mechanism is used to realize the classification and matching of wheel hub model.Through matching with the templates in the template library,the identification result of wheel hub model is obtained.The experimental results show that the identification accuracy of this scheme reaches 98.19%,and it also has a good result on the models which are easy to be confused in the traditional scheme.