A Pedestrian Re-Identification Method Based on Projection Matrix Constraints Combined with Discriminative Dictionary Learning
A pedestrian re-identification and dictionary learning technology, which is applied in the field of digital image recognition, can solve problems such as pedestrian matching difficulties, achieve good robustness, reduce time-consuming, and improve performance
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0041] Embodiment 1: as figure 1 As shown, a pedestrian re-identification method based on projection matrix constraints combined with discriminative dictionary learning, constructs a learning model that can match the perspective pictures belonging to the same pedestrian under multiple cameras, first prepare the camera acquisition under multiple perspectives Extract the features of the obtained pictures, use them as training samples, and then build a dictionary learning algorithm model, and use the projection matrix under different viewing angles as a model constraint to improve the performance of the model in distinguishing different pedestrians, and then use the training samples and model to iterate By solving the model parameters, we can obtain the sparse coding in the pedestrian training samples, and use this coding to perform similarity matching. We use angular similarity and Euclidean distance for the matching method, and give them different weights respectively. Finally, ...
Embodiment 2
[0067] Embodiment 2: as figure 1 As shown, a pedestrian re-identification method based on projection matrix constraints combined with discriminative dictionary learning constructs a learning model that can match the perspective pictures belonging to the same pedestrian under multiple cameras. Firstly, prepare pictures collected by cameras under multiple viewing angles, perform feature extraction, and use them as training samples. Secondly, a dictionary learning algorithm model is constructed, and the projection matrix under different viewing angles is used as a model constraint to improve the performance of the model in distinguishing different pedestrians. Then, relying on the training samples and the model, iteratively solves the model parameters to obtain the sparse coding in the pedestrian training samples. Finally, similarity matching is performed with this encoding. We use angle similarity and Euclidean distance for the matching method, and give them different weights. ...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More