System and method for visual tracking

Abstract

The invention discloses a system and method for visual tracking, and belongs to the technical field of visual monitoring. The method comprises the following steps: (1) collecting and inputting real-time monitoring images and target images, (2) matching the local features of the collected images and the target images, (3) finding a dense region of feature points through a clustering classification algorithm, and calculating the coordinate of a centroid where a target is located, (4) cutting out images close to the centroid, multiplying the images by a two-dimensional Gauss template, and pre-processing images to be matched, (5) carrying out overall color matching on the images to be matched and the target images in a RGB three-color channel, (6) analyzing the motion trail of the centroid, predicating a position where the centroid is likely to appear at the next moment, outputting the offset coordinate of the target relative to the center of a monitoring and shooting image in real time, and controlling a camera to track the motion trail of a suspicious target. The method can accurately find the centroid of the suspicious target and reduces the time complexity and the space complexity of target detection, meanwhile, the potential position of the target at the next moment is predicated, and real-time tracking of the suspicious target is achieved.

Description

technical field [0001] The invention belongs to the technical field of visual monitoring, and relates to a visual tracking system and method. Background technique [0002] For important monitoring places, such as banks, stores, parking lots, military bases, etc., due to the needs of management and security, people often need to detect and track suspicious objects appearing in the scene. The video-based intelligent monitoring system is a cutting-edge technology that has attracted much attention in the field of image processing, and has a huge monitoring market demand. A monitoring device for real-time detection and tracking, which can automatically analyze and extract key information of video sources, instead of simply treating the monitoring camera as an electronic device for storing and calling monitoring information. Its core processing module can use the powerful data processing of the computer Function, high-speed analysis of massive information in the video screen, fil...

AI Technical Summary

Problems solved by technology

Although the monitoring system based on the above ideas is more and more widely used in practice, there are still some unresolved problems

[0004] Problem 1: The ill-conditioned problem of 3D to 2D projection

When the camera field of view is different from the shooting angle of the template picture, the outline information of the target will be distorted, and the monitoring system using this method will not be able to detect suspicious targets

Question 2: Effects of illumination changes, shadows, occlusions, scale changes, etc.

Problem 3: The monitoring algorithm is highly complex and cannot track targets in real time

Images