Pedestrian re-identification method and system based on color texture distribution feature

Abstract

The invention belongs to the technical field of image processing, and relates to a pedestrian re-identification method based on color texture distribution features, comprising the following steps: inputting N image pairs to be matched including training data and test data and their corresponding labels 1 n , where n=1,...,N; extract the color texture spatial distribution feature representation of the input image data; obtain the consistent feature representation of the color texture spatial distribution feature representation through multi-scale feature matching; Feature representation builds a binary classifier that outputs a probabilistic representation describing the same object. The invention has the advantages of simple and fast feature extraction, short training period, high robustness and high precision.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a pedestrian re-identification method based on color texture distribution features, in particular to a deep learning method for pedestrian re-identification by a feature construction method representing image color texture distribution information. Background technique [0002] The task of pedestrian re-identification is to deal with the problem of cross-camera pedestrian matching. The application of this technology in the pedestrian monitoring network is reflected in pedestrian tracking, human body retrieval, etc. The pedestrian tracking application is to use the existing urban monitoring network to retrieve the pedestrian target to be tracked in each camera. By connecting the camera where the target appears, the movement track of the target pedestrian can be obtained, so as to complete the tracking of the target pedestrian. Human body retrieval is to retrieve the target ...

AI Technical Summary

Problems solved by technology

However, how to effectively express this information is one of the bottlenecks affecting the development of pedestrian re-identification technology.

At present, some methods at home and abroad are often based on the assumption that the pedestrians in the image remain upright, slice the pedestrian image in the vertical direction, and obtain some significant color and texture distribution features for each slice, but because of the change of the camera angle of view, the two The corresponding slices of pedestrian images do not necessarily describe the common parts of pedestrians, resulting in poor performance in complex scenes

[0004] At present, there are certain limitations and deficiencies in the construction and extraction of image texture features in the task of pedestrian re-identification: 1. Loss of spatial distribution information of color texture features

For a long time, the processing of color and texture features in the field of computer vision is often a histogram of color or texture distribution in a statistical area, and expressed in the form of a one-dimensional vector. One of the main problems caused by this method is the loss of spatial information.

2. Lack of high-level feature description methods for describing image color texture

For a long time, image-based color and texture features are often obtained directly from the original image. Due to the impact of the input image scale, the traditional feature description method lacks an effective abstraction and encoding mechanism for color and texture features, so it cannot overcome the contradiction between the curse of dimensionality and information loss.

Images