Cluster and Expansion Algorithm for Moving Object Point Cloud Based on Millimeter-Wave Radar

When using millimeter wave radar for attitude recognition of moving targets, the radar point cloud data has the characteristics of many noisy points and discrete distribution, and the traditional density space-based clustering algorithm for point cloud clustering imaging process, there will be problems such as point cloud classification error between neighboring targets and clustering of the same target point cluster into multiple point clusters. To address the above situation, a motion multi-target neighboring point cloud optimization clustering algorithm is proposed to correct the clustering results using an adaptive distance-weighted fuzzy c-mean algorithm, which improves the accuracy of near-neighbor target point cloud clustering. Meanwhile, a target point cluster expansion aggregation algorithm is proposed, which utilizes Kalman filtering for motion target position prediction, and the multi-frame iterative 3D point cloud dimensions are used as a wavegate to expand the point clusters of the target point cloud to improve the target point cloud integrity. The experimental results show that the proposed method can effectively improve the clustering accuracy.