Multi-target tracking method based on semantic information and scene information

Abstract

The invention discloses a multi-target tracking method based on semantic information and scene information, and the method comprises the steps: predicting the position of each tracking target in a scene in a current frame, carrying out the correction of the predicted position through employing a detection model, and obtaining a semantic score of the predicted position; taking the corrected position as a target frame, obtaining the similarity between the target frame and the historical track of the corresponding tracking target, and fusing the semantic score and the similarity to obtain the tracking score of the target frame; updating a scene model of the scene grid according to the tracking score of the target frame, calculating a scene confidence coefficient of the target frame accordingto the scene model, and updating the tracking score of the target frame according to the scene confidence coefficient; and obtaining a detection result of the current frame by using the detection model, matching the target frame with the detection result, and determining the state of the tracking target or generating a new target according to the matching result and the tracking score of the target frame, thereby obtaining a tracking result of the current frame. According to the invention, the robustness and accuracy of multi-target tracking can be improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and more specifically relates to a multi-target tracking method based on semantic information and scene information. Background technique [0002] Multi-target tracking is an important task in the field of computer vision, and has important application value in automatic driving, robot navigation and motion analysis. The goal of multi-object tracking is to estimate the position of all tracked objects in the scene in each frame of the image and keep the id of the same object unchanged to generate object trajectories. The existing multi-target tracking algorithms can be divided into two categories: offline algorithms and online algorithms. In offline algorithms, the multi-target tracking task is usually described as an optimization problem. By establishing a simplified model such as a network flow model, a k-part graph model or a graph cut model, an optimization algorithm is used to find t...

AI Technical Summary

Problems solved by technology

However, the existing online multi-target tracking algorithms only mechanically combine detection algorithms and single-target tracking algorithms. In fact, detection and tracking are processed separately, which will bring two problems: (1) the detection process is only for a single Image, without using the sequence information of the video, is prone to track discontinuity; (2) the tracking process does not use the semantic information used in detection, and is prone to tracking drift

Due to these two problems, existing multi-target tracking algorithms cannot overcome problems such as frequent occlusions, complex scenes, and camera motions.

Images