Human Body Identification Model for Multi-Camera Re-Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In multi-camera monitoring scenarios, such as smart retail or public places, accurately identifying individuals across images captured by cameras of different types and qualities is challenging due to varying angles and imaging quality, leading to low precision in person detection and identification.
Innovation Solution
A method involving a trained human body identification model that performs feature extraction on input images from diverse cameras and matches these features with a preset database to identify individuals, using a feature extracting module and a determining module within a machine learning model, which is iteratively adjusted to improve identification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple cameras of different types are used to expand imaging scope, then coverage area is improved, but identification accuracy deteriorates due to varying angles and imaging quality
Solution Approach 1:
The patent creates a unified identification system that processes images from multiple camera types (ceiling-mounted, wall-mounted, portable) with different functions and qualities. The system extracts features from diverse image sources and integrates them into a single identification framework, allowing the system to handle various camera inputs universally while maintaining accurate person identification across all imaging conditions
Solution Approach 2:
The patent transforms images from different camera types by adjusting parameters such as resolution, orientation, and feature extraction thresholds. The system dynamically modifies processing parameters based on the specific camera source and image quality, converting diverse inputs into a standardized format suitable for accurate comparison and identification
2Productivity
If feature extraction is performed on images from cameras with different imaging qualities, then comprehensive person detection is improved, but matching precision deteriorates
Solution Approach 1:
The patent applies different feature extraction strategies to different regions and image qualities. For high-quality images, detailed facial features are extracted, while for lower-quality images from ceiling or wall cameras, the system focuses on distinctive body features such as clothing patterns, posture characteristics, and gait. This localized approach ensures that each image contributes maximally to identification without compromising overall precision
Solution Approach 2:
The patent introduces an intermediate feature representation layer that bridges images from cameras with different qualities. This intermediary layer normalizes features from various sources before final comparison, acting as a mediator that harmonizes the differences in imaging quality while preserving essential identification information for accurate matching
Data Source
AI summary
Embodiments of the present disclosure disclose a method, electronic device, and computer readable medium for image identification. The method comprises: acquiring an image comprising a person object for use as an input image; performing feature extraction on the input image using a feature extracting module of a trained human body identification model; and matching an extracted human body feature of the inputted image with a preset human body feature database, to identify the person object in the inputted image, wherein the human body identification model extracting features of human body images captured by cameras of different categories respectively using the feature extracting module, and the human body identification model identifying whether the human body images captured by the cameras of different categories are human body images of a given person based on the extracted features of the human body images. This method improves the accuracy of multi-camera human body re-identification.