On-line tracking method for designated pedestrian under a cross-camera scene

Abstract

The invention relates to an on-line tracking method for designated pedestrian under a cross-camera scene, which can create and initialize a tracker according to the obtained basic information by demarcating a pedestrian of interest in the visual field area of any camera; when a pedestrian is appointed to leave the view area of the current camera, the pedestrian state is set as a hanging detection,waiting for the next step of pedestrian data correlation operation; when a new pedestrian enters into the tracking visual field area, judge whether the pedestrian in the hanging state and the new target pedestrian is the same pedestrian through an association algorithm; according to the data correlation algorithm based on depth learning, similar values of two target pedestrians are obtained, andwhen the similarity value is greater than a threshold value, the corresponding tracker is updated to obtain a complete movement track of the target pedestrian of interest in the whole camera network.The invention provides an on-line tracking method for designated pedestrian under a cross-camera scene, and the practicability and the accuracy of the target tracking of a non-overlapping field-of-view area under a multi-camera scene is increased.

Description

technical field [0001] The invention relates to the field of intelligent security, in particular to an online tracking method for designated pedestrians in a cross-camera scene. Background technique [0002] With the continuous development of the economy, people's demand for safety is increasing day by day. Therefore, the application range of intelligent surveillance video in the field of security is constantly expanding, and related video intelligent processing technologies such as pedestrian tracking and retrieval have become the focus of current research. Its research scope extends from the traditional technology only for single-camera application scenarios to multi-camera application scenarios, and the research object has evolved from the previous single pedestrian to the need to process multiple pedestrians at the same time. At present, the pedestrian tracking technology under a single camera is relatively mature, but the existence of blind spots in multi-camera, espec...

AI Technical Summary

Problems solved by technology

For most of the current cross-camera pedestrian tracking algorithms, the ability to learn pedestrian features is limited, and more robust pedestrian features cannot be learned.

Therefore, it ultimately affects the accuracy of the subsequent pedestrian similarity measurement, which eventually leads to unsatisfactory data association results.

So it is difficult to adapt to the complex environment of cross-camera pedestrian tracking

[0005] Although the existing cross-camera pedestrian tracking related research can effectively solve some pedestrian tracking for offline data, it cannot meet the requirements for real-time online tracking, and it cannot be effective when unknown pedestrians enter and exit the area. track

Images