Object tracking method with minimum trajectory entropy

Abstract

The invention provides an object tracking method with minimum trajectory entropy, and belongs to the technical field of computer graphs and images. The method specifically comprises the following steps of: selecting a target, wherein a target to be tracked is selected and determined from an initial image; inputting images, wherein in a real-time processing condition, video images collected by a camera and stored in a storage area are extracted as input images for tracking, and in an off-line processing condition, the collected video file are decomposed into an image sequence formed by a plurality of frames, and the frame images are extracted according to a time sequence as input images; creating detectors whose number is no more than 5; generating a tracking trajectory, wherein all the detectors simultaneously detects the target; positioning a trajectory entropy minimum target; updating the detectors, wherein the detector training process in the step (3) is resembled according to the determined target trajectory, positive and negative samples are respectively extracted for each frame, corresponding to the target trajectory, in a target area and a background area, and all detectors are updated. The object tracking method is used for a video monitoring system.

Description

technical field [0001] The invention belongs to the technical fields of computer vision, graphics and images. Background technique [0002] Visual object tracking is one of the most important components in computer vision applications, such as intelligent surveillance, human-computer interaction, automatic control systems, etc. The purpose of object tracking is to automatically determine its position and size in each subsequent frame given the position and size of the initial object. Although the research on object tracking has been carried out for decades, and a lot of important progress has been made in recent years, due to the complexity of the real world, such as background interference, appearance and illumination changes, low image quality, frame skipping, etc. Designing tracking methods that can achieve human-level performance is still very difficult. An ideal tracking method must consider the real-time performance, stability and persistence of tracking at the same ...

AI Technical Summary

Problems solved by technology

However, these methods are still difficult to adapt to more complex tracking environments, which provides a possible solution to the analysis and processing of tracking trajectories.

Images