Target detection method, intelligent device and computer readable storage medium
By acquiring historical detection results of the current data frame and performing N rounds of attention processing in the target detection method, the problem of high computational cost in traditional methods is solved, achieving low-cost and accurate target detection, with computational cost not increasing with the time window.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 安徽蔚来智驾科技有限公司
- Filing Date
- 2025-01-17
- Publication Date
- 2026-07-17
AI Technical Summary
Existing target detection methods involve high computational costs when using continuous multi-frame data collected by sensors, especially as the time window increases, making it impossible to effectively utilize temporal information for accurate target detection.
By acquiring the current data frame collected by the sensor on the smart device and the historical detection results within the previous preset time window, the target detection model performs N rounds of attention processing to obtain the current detection result and stores it in the preset memory, thereby reducing the amount of computation.
While maintaining low computational costs, it achieves accurate target detection based on the temporal information of continuous multi-frame data collected by sensors. The computational load does not increase significantly with the increase of the time window and is close to that of a single-frame detection model.
Smart Images
Figure CN122415971A_ABST