A pedestrian searching method and device based on structural perception self-attention and online instance aggregation matching

Abstract

The invention discloses a pedestrian searching method and device based on structural perception self-attention and online instance aggregation matching, and belongs to the technical field of computervision technology processing. The method comprises the following steps: firstly, in a training phase, combining a convolutional neural network with a non-local layer; carrying out feature extraction on an input whole scene image to obtain feature representation of the scene image, designing structure-perceived anchor points for a special object of a pedestrian, improving the performance of a detection framework, framing the detected pedestrian into the same size, then sending the pedestrian into a pedestrian re-identification network, and carrying out training, storage, optimization and updating of pedestrian features with tags. In the model testing stage, the trained non-local convolutional neural network is used for carrying out pedestrian detection on an input scene image, and after a pedestrian frame is detected, a target pedestrian image is used for carrying out special similarity matching sorting and retrieval. Pedestrian detection and re-identification can be carried out on large-scale real scene images at the same time, and the method plays an important role in the security and protection fields of urban monitoring and the like.

Description

technical field [0001] The invention belongs to the technical field of computer vision technology processing, and further relates to a pedestrian search method based on structure-aware self-attention and online instance aggregation and matching in the technical field of target detection and target retrieval. Background technique [0002] The document "Joint detection and identification feature learning for personsearch, Computer Vision and Pattern Recognition (CVPR), 2017IEEE Conference on.IEEE, 2017: 3376-3385." discloses a new framework for pedestrian search that integrates pedestrian detection and pedestrian re-identification. The current pedestrian re-identification benchmarks and methods are mainly to match the cropped pedestrian pictures, but the real scene is not so ideal. When doing pedestrian search, it is necessary to use the pedestrian detection method to mark pedestrians first, and then use pedestrian re-identification method to search for a specific person. [...

AI Technical Summary

Problems solved by technology

In the method described in the literature, when detecting pedestrians, problems such as false positives, missed detection, and misplaced bounding boxes will inevitably occur, which will affect the effect of pedestrian search. The limitations of the convolutional neural network prevent the model from learning a global The distributed information cannot locate the areas with dense pedestrians well. When the overall scale of the data is small, the image information such as people's gestures, behaviors and actions is not rich enough, and there are few pedestrian samples of the same label, a single online instance The matching loss function cannot make the model learn discriminative features

Images