Target recognition algorithm based on sparse LiDAR scanning data

Abstract

For target recognition of sparse LiDAR scan data, this paper proposes a new method for aligning height data matrices based on rotation, length segmentation, and grid partitioning. Initially, the method involves rotating the matrix to achieve initial pose alignment between the height matrix of the target to be recognized and the template matrix. Subsequently, utilizing length segmentation, the orientation of the target is aligned, and the initial grid search position is determined. Finally, the method searches for the best alignment grid based on grid partitioning to achieve data matrix alignment and performs target recognition using a fuzzy recognition algorithm based on scanlines. Experimental results demonstrate that this method effectively aligns sparse target data with target template data, also enabling the recognition of typical targets such as cars, providing a valuable reference for target recognition of sparse LiDAR scan data.