Measurement learning method and system for pedestrian re-recognition

Abstract

The invention discloses a measurement learning method and system for pedestrian re-recognition. Implementation of the method comprises the steps of collecting feature vectors of a pedestrian target under two cameras, building a positive sample pair feature vector set and a negative sample pair feature vector set, calculating the distance of positive sample pair feature vectors and the distance of negative sample pair feature vectors, restraining the negative sample pair feature vectors by adopting dual thresholds, building a measurement matrix based loss function under distance constraint conditions of the positive and negative sample pair feature vectors, and iteratively updating the measurement matrix by taking the minimum loss function value as the target. The measurement learning method can effectively reduce influences imposed on matrix learning by irrelevant variables such as image background and noise, so that occurrence of an overfitting phenomenon is avoided, and the acquired measurement matrix is high in generalization ability.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition, and more particularly relates to a metric learning method and system for pedestrian re-identification. Background technique [0002] Pedestrian re-identification algorithm is one of the important fields of image processing and pattern recognition research, focusing on the recognition of specific target pedestrians under cameras without public view. Due to limitations such as video clarity, it is difficult to find the same target through intuitive information such as faces. Instead, we focus on the feature expression based on the appearance of pedestrians, mainly including information such as the color and texture of pedestrian images, and then find a suitable measurement. The method makes the cross-camera features of the same pedestrian target as similar as possible, and the difference of heterogeneous target features is as significant as possible. Since the same target is affected b...

AI Technical Summary

Problems solved by technology

[0003] The present invention proposes a metric learning method for pedestrian re-identification. The purpose is to provide a metric learning method that imposes single-threshold constraints on the feature vectors for positive samples and double-threshold constraints for negative samples on feature vectors. Overfitting problem in complex scenes

Images