Perception decision integration target detection and tracking method based on high and low perspective cooperation
By employing network time synchronization, CSPDarknet feature extraction, homography correction and geographic coordinate registration, unified geospatial metric, and multi-branch spatiotemporal graph convolutional network, the problems of perspective distortion and latency in the air-ground collaborative monitoring system were solved. This achieved high-accuracy entity re-anchoring and low-latency perception-decision consistency, thereby improving the cross-view target tracking capability of air-ground cameras.
Patent Information
- Application Number
- CN202610669005.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-15
- Publication Date
- 2026-07-24
- Estimated Expiration
- 2046-05-15
AI Technical Summary
Existing technologies in air-ground collaborative monitoring systems suffer from problems such as trajectory chain breakage caused by severe perspective distortion, system crashes caused by dead point overflow in reverse perspective mapping, and high response latency caused by the fragmentation of the perception and decision-making link. These issues make it difficult to achieve cross-perspective area mapping, accurate target tracking, and edge-decentralized state fusion.
It adopts high and low bit video stream access and temporal multimodal alignment based on network time synchronization protocol, combines target detection and deep feature extraction of CSPDarknet backbone network, cross-view spatial mapping based on homography correction and geographic coordinate registration, uses trajectory association integration with unified geospatial distance metric and bipartite graph matching, and realizes integrated generation of perception and decision through multi-branch spatiotemporal graph convolutional network ST-GCN.
It achieves uninterrupted, high-accuracy entity re-anchoring under conditions of rapid elevation and depression angle changes, solves the problems of feature collapse and soaring matching and recognition error rates in traditional systems, realizes a low-latency perception-decision consistency closed loop, and improves the cross-view area mapping anti-distortion and target accuracy tracking capabilities of air-ground camera collaboration.