Specific person re-identification method based on multi-source image information

Abstract

The invention discloses a specific person re-identification method based on multi-source image information. The method comprises the steps: setting a plurality of fixed cameras and unmanned aerial vehicles with pan-tilt cameras in the same scene, and setting an image cache queue to perform video acquisition and storage synchronization on the plurality of cameras; performing real-time pedestrian detection on images in the video by adopting a first-stage target detection network, and storing the images in a cache region; constructing a pedestrian re-identification network for processing to obtain re-identified Euclidean distance, and drawing and displaying a personnel position frame after sorting. According to the method, the image cache queue is set by evaluating the efficiency of the neural network algorithm, the monitoring video image can be effectively and stably acquired in an actual scene, the method can adapt to different deep learning algorithms to improve and perfect the scheme, and the re-recognition efficiency is improved.

Description

Technical field [0001] The present invention belongs to the field of video surveillance, and more particularly to a specific person reconciliation method based on multi-source image information. Background technique [0002] Pedestrians reintegrate the computer visual field in recent years, a research topic, can be considered a child problem of image retrieval, and its goal is to give a surveillance pedestrian image to retrieve the pedestrian image under other devices. The traditional method relies on manual features that cannot be adapted to a large amount of data. In recent years, with the development of deep learning, a large number of pedestrian re-identification methods based on deep learning is proposed. [0003] According to the network training, the method of the pedestrian re-identification can be divided into two categories of characterization and metrics. Character learning and metrics have their own advantages and disadvantages. At present, the academic community and ...

AI Technical Summary

Problems solved by technology

[0007] Some current pedestrian re-identification methods still have the problem of being unable to perform real-time processing for video surveillance. If real-time processing cannot be completed, pedestrian re-identification will easily cause a backlog of unprocessed surveillance videos when applied to actual scenarios, which will eventually cause storage or There is a problem with the re-identification result (such as delay, memory overflow, etc.)

Therefore, it is difficult to apply these methods to practical scenarios.

If the surveillance video is processed through multiple identification devices, the hardware cost will undoubtedly be greatly increased

In addition, the number of surveillance videos in the actual scene is limited and often cannot cover a specific area. In this case, the difficulty of re-identifying and positioning a specific person is greatly increased.

Images