Person Re-identification Based On Salience Features

In the multiple cameras monitoring system, the research of person re-identification faces a big challenge because of the different camera views, the change of the background, different scales and some other force majeure factors. The problem of person re-identification is desperately to be solved because that the features obtained is small and the discrimination is not big enough especially in the low resolution images. In order to solve these problems and to promote the development of the intelligent monitoring technology, this paper proposes person re-identification algorithm based on salience features. Using the salience features of images for matching improves the robustness and accuracy of the algorithm.In this paper, the image of a person is divided into patches of the same size, and the features are extracted based on these patches. This approach well solves the problem caused by misalignment or different camera views. The features used are LAB color histogram and SIFT feature which is represented by a discriminative descriptor vector. A dense correspondence is implemented based on the salience values. When calculating the salience value, the combination of between class salience value and the intrinsic salience in the original image is used to better describe the salience features of the patches.