Human Body Recognition via 3D Back-Projection Error Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current multi-person tracking and recognition systems using two-dimensional images across multiple cameras suffer from significant variations in human body posture, leading to low accuracy and recognition errors due to differences in visual features.
Innovation Solution
Introducing three-dimensional spatial coordinates into human body recognition technology to pre-judge recognition results and perform re-recognition using person re-identification ReID, calculating back-projection errors, and re-recognizing images until all cameras' errors are within a preset threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If two-dimensional image information is used for multi-person tracking and recognition across multiple cameras, then the system can operate with simpler processing, but the recognition accuracy deteriorates due to large posture variations and visual feature biases
Solution Approach 1:
The patent transforms the recognition problem from two-dimensional image space to three-dimensional space by calculating back-projection errors. This dimensional transition allows the system to account for posture variations and camera angles, thereby improving recognition accuracy without significantly increasing overall system complexity
Solution Approach 2:
The patent implements an iterative re-recognition process where back-projection errors are calculated and used as feedback to adjust and refine recognition results. This feedback mechanism continuously improves accuracy by comparing predicted and actual positions across multiple cameras
2Measurement precision
If three-dimensional spatial coordinates are introduced for human body recognition, then recognition accuracy is improved by accounting for posture variations, but the calculation and processing complexity increases
Solution Approach 1:
The patent introduces three-dimensional spatial coordinates to capture human body pose information, transforming the recognition problem from 2D image space to 3D space. This enables the system to model posture variations and camera angle effects, significantly improving recognition accuracy
Solution Approach 2:
The patent replaces complex visual feature matching mechanisms with a geometric back-projection error calculation approach. By using camera matrices and coordinate transformations, the system substitutes intricate image processing with more efficient geometric computations, reducing overall processing complexity
3Reliability
If back-projection error calculation is performed for each camera, then recognition errors can be detected and corrected, but the computational time and processing load increase
Solution Approach 1:
The patent calculates back-projection errors for each camera and uses these errors as feedback to identify and correct recognition mistakes. This feedback mechanism iteratively refines recognition results, improving reliability by continuously comparing and adjusting predictions against actual observations
Solution Approach 2:
The patent performs back-projection error calculations selectively for cameras where recognition errors are detected, rather than uniformly processing all cameras. This partial action approach optimizes processing time by focusing computational resources only where needed, balancing reliability improvement with efficient resource utilization
Data Source
AI summary
The present disclosure provides a human body recognition method and apparatus, and a storage medium, the method comprising: determining a coordinate of a target person in a three-dimensional space according to images containing the target person collected by at least two cameras; calculating back-projection errors of the target person under different cameras respectively according to the coordinate of the target person in the three-dimensional space; determining whether the cameras have a human body recognition error according to the back-projection errors of the cameras; when a camera has the human body recognition error, performing re-recognition of the target person under the camera by using person re-identification ReID, until the back-projection errors of all the cameras containing the target person are not greater than a preset threshold. The present disclosure can improve accuracy of the human body recognition result effectively.