Human body detection and tracking method and device based on unmanned aerial vehicle mobile platform

Abstract

The invention discloses a human body detection and tracking method and device based on an unmanned aerial vehicle mobile platform and belongs to the computer vision field. The technical key points of the method comprise a human body detector training step, a target human body recognizer offline training step, a target human body detection step and a human body tracking step, wherein the target human body detection step is characterized by receiving a current video frame shot by an unmanned aerial vehicle, extracting object characteristic value in the current video frame, sending the object characteristic value to a human body detector, the human body detector judging whether a human body is detected according to the characteristic value, and if so, further sending the characteristic value to a target human body recognizer, the target human body recognizer judging whether a target human body is detected according to the characteristic value, and if so, labeling the characteristic value and adding the characteristic value to a tracking list; and the human body tracking step is characterized by predicating position of the target human body in the next video frame according to the coordinate position of the target human body in the current video frame.

Description

technical field [0001] The invention belongs to the field of computer vision, and is a real-time target tracking method, especially a long-term and stable human body tracking method under the complex background of an unmanned aerial vehicle platform. Background technique [0002] Unmanned aircraft, referred to as "UAV" or "UAV" in English, is an unmanned aerial vehicle controlled by radio remote control equipment or its own program control device. Among the many application fields of UAVs, target tracking, especially the tracking of human body, has always been a difficult problem to be solved urgently. This is because the human body usually has a small image in the field of view of the optical imaging device of the drone, the resolution is not high, and it is non-rigid and irregular, and its appearance and outline change at any time; and the tracking target is in a complex ground scene In the background, there are problems such as poor contrast, occlusion, and stereoscopic ...

AI Technical Summary

Problems solved by technology

This method uses the traditional MeanShift algorithm and Kalman filtering method to track the target. When the target is blocked or jumps, this method cannot guarantee long-term stable tracking.

[0004] In summary, in the application of UAVs, there is currently no long-term, stable, and anti-occlusion tracking method for moving targets, especially non-rigid human targets.

Images