Adaptive Feature Segmentation Method In Assembly Accuracy Detection And Its Application On Virtual Assembly

Assembly precision detection is the key to ensure the quality of assembly and performance, widely applicated in assembly production, and in a virtual environment using the fitting precision of point cloud model and ideal model matching the fitting precision of calculation is a kind of common testing methods.Feature segmentation can be virtual environment are obtained through automation assembly with characteristic information, only access to the information under the environment of virtual assembly accuracy calculation, so the automation features segmentation is the key link in virtual assembly accuracy of detection. In order to in a virtual environment to test the assembly accuracy, need to be able to reflect the real shape, size, and assembly parts tooling information such as error of the model.Point cloud model, however, is only the actual 3 d coordinate data collection of parts surface, and no assembly with characteristic information, so the lack of automation feature segmentation method can lead to difficult to test the virtual assembly precision problems.The first chapter introduces the assembly precision detection technology research at home and abroad, including virtual assembly accuracy detection technology, point cloud model features segmentation technology, intensive data clustering technology, analyzes the research progress of the technology and the existing problems, thus put forward in this paper, the research content and analyze the research significance, and finally simply introduces the structure of the thesis.The second chapter puts forward the point cloud model based on random sample consistency characteristic segmentation algorithm.First of all, the most significant feature is the components of space characteristics, it is the main difference in chamfering, narrow face such details characteristics, assembled by the point cloud model with the characteristics of the parametric definition, from the point of view of statistical and functional analysis of the random sample consistency feature segmentation method of characteristics.Then, according to the definition and analysis of the above, design the model of point cloud based on random sample consistency characteristic segmentation algorithm, and carries on the experimental analysis.The third chapter is put forward based on the weight distribution of soft k-means clustering analysis of RANSAC detail feature segmentation algorithm.First of all, the detail feature is the component of chamfering, narrow face such small local characteristics, the characteristics of the analysis of these data in point cloud model, according to these characteristics of point cloud model for data preprocessing.Then, combining with clustering analysis theory in unsupervised learning, small local characteristics developed in view of the point cloud model of clustering.Finally, using the clustering of RANSAC segmentation algorithm was improved and integrated form some adaptive characteristics of cloud model segmentation method.The fourth chapter puts forward the use of point cloud model and error associated mixed model assembly accuracy detection method to compute the ICP registration.First use of characteristics of point cloud model adaptive segmentation method to obtain the actual assembly of parts assembly feature information, and then point cloud model mixing model error associated with the position after the ICP registration information, assembly accuracy evaluation based on LOOCV optimization calculation, so as to realize the assembly accuracy of detection of virtual environment.The fifth chapter assembly accuracy testing system is developed, realizing virtual environment model of point cloud data preprocessing module, assembly precision testing flow control module, the point cloud model adaptive feature segmentation module, assembly precision detection calculation function module.And the system is applied to actual parts assembly accuracy of detection.The sixth chapter summarizes the full text, and analyses our study on deficiencies and projections for the future research work and prospects.