复杂动态环境下点云时变追踪建模方法、装置、设备及介质
By constructing a spatiotemporal point cloud sequence and using feature fusion technology, the problem of noise interference in point cloud tracking under complex dynamic environments was solved, achieving high-precision target tracking and avoiding trajectory drift and target loss.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN UNIV
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-17
AI Technical Summary
In complex dynamic environments, the 3D point cloud flow contains a large number of false targets and discrete noise points formed by suspended dust and water mist reflection, which leads to trajectory drift or target loss in target tracking using traditional methods.
By constructing a spatiotemporal point cloud sequence, selecting key point cloud sequences, extracting target features and reference features for fusion, and using multilayer perceptron and multi-head attention mechanism for feature interaction, the target position is predicted and adjusted, discrete noise is eliminated, and the device skeleton structure is preserved.
It improves the accuracy of target tracking, avoids trajectory drift or target loss caused by noise points, and achieves high-precision tracking in complex environments.
Smart Images
Figure CN122089785B_ABST
Abstract
Citation Information
Patent Citations
Mesh based frame processing and applications
CA2466247A1
Feature fusion method, system and device based on visual radar, medium and product
CN120088606A