A Multi-modal Point Cloud Data Preprocessing Approach by Integrating Template Matching and Improved Parameter-free Filtering

Abstract

Abstract Point cloud preprocessing plays an important role during measuring sizes of workpieces or parts by using point cloud data, which directly impacts the accuracy of subsequent point cloud registration. The exisiting preprocessing methods primarily concentrate on single-source data processing, which often lead to some problems, such as low performance, high computational demands, and lengthy adjustment times, so that they cannot be applied to measure the sizes of industrial parts. Therefore, a multi-modal point cloud data preprocessing approach by integrating template matching and improved parameter-free filtering is presented in this paper. A spatial transformation and template matching method are firstly employed to integrate 2D image data with 3D point cloud data of industrial parts, facilitating the extraction of point cloud features. An improved parameter-free filtering approach is subsequently proposed to eliminate some point cloud clusters generated by residual background edges that could not be eliminated even after preprocessing with the template matching method. Experiment results demonstrate the effectivency of the aforementioned methods.

Affiliated Institutions

• Mongolian University of Science and Technology MN
• Micron (United States) US

Related Publications