A cross-camera re-identification fusion method and system for objects with similar appearance

Abstract

The present invention provides a cross-camera re-identification fusion method for objects with similar appearances, which uses a deep convolutional neural network to extract the global feature map of the picture, and extracts the appearance vector of the target on the global feature map according to the target detection result; encodes the camera, Generate a view vector containing observation view angle information; generate a position vector of the target according to the position of the detection frame corresponding to the target in the image coordinate system. The three vectors are fused and transformed to generate the target representation vector. The network is trained by optimizing the triplet loss function, and the representation vector for re-identification is learned. During the training process, a combination of offline mining and online mining is used to generate and update the triplet dataset. Finally, the hierarchical clustering algorithm with constraints is used to cluster the representation vectors corresponding to the targets in different cameras to realize cross-camera target re-identification. At the same time, a cross-camera re-identification fusion system for objects with similar appearance is provided. The invention improves the accuracy of re-identification.

Description

technical field [0001] The present invention relates to a re-recognition technology in the technical field of pattern recognition, in particular to a cross-camera re-recognition fusion method and system for objects with similar appearances. Background technique [0002] Target re-identification refers to: for a certain or some given targets, determine whether there is a target with the same identity in other times or perspectives. Re-identification was first proposed in the field of multi-camera tracking of targets, and it is also one of the key technologies for multi-camera tracking. Specifically, when the same target appears under the perspectives of multiple cameras, it is necessary to determine the uniqueness of the target identity observed by multiple observers. In terms of timing, the identity of the target's identity at different moments also needs to be maintained by re-identification technology. Object re-identification, especially pedestrian re-identification, ha...

AI Technical Summary

Problems solved by technology

However, the appearance of the target is often affected by many external factors, so this method has great instability

In addition, for some objects (such as types of vehicles), the appearance of individuals is relatively similar and the difference is small, which often makes the appearance information invalid when distinguishing them, making it difficult or impossible to achieve re-identification through the appearance model

At present, there are few researches on re-identification of such targets with similar appearance, which belongs to blank field

Images