Target tracking method and system based on straight line edge feature

Abstract

The invention provides a target tracking method based on straight line edge features. The target tracking method comprises the step of acquiring the straight line edge. The step of acquiring the straight line edge comprises the following steps that step 1: the image vertical edge and the horizontal edge are acquired to be detected according to the image acquired from the camera so as to acquire afirst straight line edge; and step 2: the pixel number in the connected domain of the image and the preset pixel number are compared according to the image acquired from the camera so as to acquire asecond straight line edge. According to the target tracking method based on the straight line edge features and the target tracking system based on the straight line edge features, three edge detection methods are used for providing sufficient number of edge features for the subsequent tracking task and then integrated to remove the repeated part and connect the broken edges so that target tracking can be greatly simplified.

Description

technical field [0001] The present invention relates to the field of target tracking methods, in particular to a target tracking method and system based on straight line edge features. Background technique [0002] Video-based object tracking is a research hotspot in the field of computer vision, and it has extremely wide applications in all walks of life. Common tracking methods use key points or the global features of the target to match the front and back frames. These methods tend to be more complex. [0003] On some special tracking tasks, the tracking strategy can be simplified. For example, when tracking man-made rigid objects such as ships and vehicles, you can make full use of the straight edges on the target for matching. In this way, the tracking strategy can be greatly simplified. [0004] The patent document application number is: 201610032163.7; the publication number is CN105608711A, which discloses a video dynamic target tracking and extraction method bas...

AI Technical Summary

Problems solved by technology

[0007] On the one hand, the background models of these algorithms require a large amount of video frame image training to be initialized, that is, they cannot be used immediately when the application is started, which will cause certain missed detections; on the other hand, the performance and accuracy of these algorithms in foreground area extraction Not high, there is a possibility that the head area is determined as the background area, so that the subsequent histogram of oriented gradient (Histogram of Oriented Gradient, HOG) feature cannot scan the head determined as the background area, resulting in missed detection

Images