Multi-target tracking method applied to panoramic video and panoramic camera

Abstract

The invention provides a multi-target tracking method applied to a panoramic video and a panoramic camera. The method comprises the steps that the panoramic image of a certain frame in the panoramic video is selected as the panoramic frame of an initial frame; N projection images are selected from the panoramic frame of each frame after the panoramic frame of the initial frame; M track targets areselected from all projection images corresponding to the panoramic image of the initial frame and the panoramic images of the subsequent adjacent H frames; the panoramic image of the initial frame isused as the panoramic image of the first frame; for the panoramic image of each frame after the panoramic image of the first frame, each track target in N projection images in the panoramic image ofthe i-th frame is used for tracking to acquire at least one successfully tracked target in N projection images in the panoramic image of the (i+1)-th frame. According to the multi-target tracking method applied to the panoramic video and the panoramic camera, a number of track targets can be effectively can tracked, and the tracking instantaneity of a number of track targets can be improved.

Description

technical field [0001] The invention relates to the technical field of panoramic video, in particular to a multi-target tracking method and a panoramic camera applied to panoramic video. Background technique [0002] Panoramic video is composed of some time-related panoramic image sequences, from which a real panoramic scene can be obtained, and video frames can be positioned through time. The global motion of the cube panorama video of a still scene is usually dominated by horizontal motion, covering a 360-degree scene range. For a continuous multi-frame panorama image, the target object will move from one face of the cube panorama to the adjacent face. Due to its dynamic background, it is not suitable for the detection and tracking of dynamic objects using the method of background difference, so it is necessary to design a fast and efficient object tracking method in cube panoramic video. [0003] In the existing methods for tracking objects in panoramic video, the panora...

AI Technical Summary

Problems solved by technology

In the tracking process, the model of the target object is established, and the probability of the tracking target appearing in different areas is established according to the front and rear frames, and the position with a higher probability is used as the new position of the target; the premise of using the above method is that there will be no missed detection, and the The direction of must contain the target that needs to be tracked, and in the case of missed detection, the tracking accuracy is extremely low, and when multiple targets need to be tracked, the missed detection will be greatly improved, which leads to the existence of existing methods. Panoramic video tracking multiple targets with low accuracy

[0004] And adopt above-mentioned method, firstly need to obtain panoramic video image first and then carry out projection to obtain the projected image of preset field of view, then carry out the probability that the target appears in different regions before and after the frame, determine the new position of tracking target with the magnitude of probability, its There are many calculation steps and a large amount of calculation, and when multi-target tracking is performed, the calculation amount will increase exponentially, which leads to the problem of low tracking efficiency of panoramic video

Images