Pedestrian Re-Identification with Head-Shoulder Features in Low Light
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pedestrian re-identification models rely heavily on clothing colors and styles, leading to degraded performance when these attributes cannot be distinguished, such as in low-light or all-black scenarios.
Innovation Solution
A method that extracts both global and head-shoulder features from target images, using deep learning models like ResNet50 and ResNet18, to determine a representation feature for each person, enabling accurate re-identification even in challenging conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing pedestrian re-identification models rely on clothing colors and styles, then identification can be performed under normal conditions, but identification performance degrades severely when clothing attributes cannot be distinguished
Solution Approach 1:
The patent divides the pedestrian recognition task into two independent feature extraction branches: global feature extraction (processing the entire image) and head-shoulder feature extraction (processing a cropped region). This segmentation allows the system to leverage both overall appearance and distinctive local features, ensuring reliable identification even when clothing attributes are indistinguishable.
Solution Approach 2:
The patent applies local quality by focusing on the head-shoulder region, which contains more distinctive and stable identification features compared to general clothing. By extracting features from this specific local area and combining them with global features, the system maintains high identification performance under challenging conditions where clothing colors and styles cannot be distinguished.
2Measurement precision
If only global features are used for pedestrian re-identification, then the process is simple, but accuracy degrades when clothing features are indistinguishable
Solution Approach 1:
The system segments the feature extraction process into two parallel pathways: a global feature extraction network processing the entire pedestrian image, and a head-shoulder feature extraction network processing a cropped region. This segmentation enables the system to achieve higher accuracy by combining multiple feature types while maintaining a relatively simple overall architecture through modular design.
Solution Approach 2:
The patent merges global features and head-shoulder features into a unified representation by concatenating their feature vectors. This combining strategy allows the system to leverage complementary information from both feature types, achieving superior re-identification accuracy without significantly increasing system complexity.
3Reliability
If head-shoulder features are extracted in addition to global features, then identification accuracy improves in challenging conditions, but computational complexity increases
Solution Approach 1:
Instead of processing the entire image at high resolution or using overly complex models, the patent applies partial action by extracting and processing only the head-shoulder region separately from the global image. This selective processing reduces computational energy requirements while still capturing critical identification features, thereby maintaining reliability without excessive computational cost.
Data Source
AI summary
A method and an electronic device for pedestrian re-identification are provided. The method may include: collecting a target image set including at least two target images, where each target image includes at least one person; extracting a global feature and a head-shoulder feature of each person in each target image in the target image set, where the global feature is an overall appearance feature, and the head-shoulder feature is a feature of a head part and a shoulder part determining a representation feature of each person in each target image based on the global feature and the head-shoulder feature of each person in each target image; and determining a given person in different target images based on the representation feature of each person in each target image.