Method for tracking target object
A target and object technology, applied in the field of tracking, can solve problems such as simple algorithms, features being affected by the background, and complex TLD calculations, and achieve the effect of strong practicability and full-automatic tracking
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Publication Date
- 2018-06-29
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
technical field
[0001] The invention relates to the technical field of tracking, in particular to a method for tracking an object. Background technique
[0002] Traditional vision-based target tracking algorithms include motion detection, MeanShift, KCF, TLD, etc. The motion detection algorithm is simple and cannot detect scenes with background changes; MeanShift and KCF have a small amount of calculation, and the tracking effect is good for situations where the target size does not change much. , TLD is complex to calculate. Like KCF, it has requirements for the size change of the target. For fast and small target objects, due to the small size, the features are easily affected by the background, and the tracking effect is poor. KCF and TLD are a kind of The online learning algorithm needs to manually specify the object to be tracked before starting the track. It cannot achieve automatic tracking, and it needs to be re-specified after the target is lost. The practical effec...
Examples
Embodiment 1
[0046] Please refer to figure 1 , Embodiment 1 of the present invention is:
[0047] A method for tracking a target, comprising the following steps:
[0048] S1. Pre-acquisition of multiple frames of images at one position, performing average calculation on the gray value of the same pixel in each frame of images, and calculating the first image, which is marked as a background image. This embodiment collects 20 frame image;
[0049] Collecting a frame of images at the one position, detecting moving objects in the one frame of images, and obtaining a set of moving objects;
[0050] Comparing the one-frame image with the background image, judging whether there is an object in the one-frame image that does not exist in the background image, and if so, the object is a moving object;
[0051] judging whether there is the same object whose vector displacement difference between the one frame image and the background image is greater than the first preset value, if so, judging th...