Multi-Camera Person Recognition via Object Information Matching

The patent introduces object information (such as carried objects, clothing items, or accessories) as an intermediary element to辅助 person recognition. Instead of relying solely on direct visual analysis of the person which varies across cameras, the system uses detected objects associated with the person as a stable reference point for identification across different camera views, thereby improving reliability without sacrificing precision

The system employs a multi-functional recognition approach that combines both direct person visual analysis and object information analysis. This universal method works across different camera types and viewing angles by leveraging multiple data sources (person appearance, carried objects, clothing) to achieve reliable and precise identification in diverse monitoring scenarios

Object information serves as a mediator that simplifies cross-camera person matching. By detecting and tracking objects associated with persons (such as shopping carts, bags, or distinctive items), the system creates a common reference framework that works across multiple cameras regardless of viewing angles and lighting variations, reducing the complexity of matching persons across the expanded monitoring area

The system adds an additional dimension to person identification by incorporating object information as a separate feature space. Instead of relying solely on 2D visual appearance which varies across cameras, the system uses object detection results as an independent dimension for matching, creating a more robust multi-dimensional identification approach that handles multi-camera complexity