Unlock instant, AI-driven research and patent intelligence for your innovation.

Fast Moving Target Tracking Method Based on Kalman Target Prediction and Multi-feature Compression Fusion

A target tracking and target prediction technology, applied in the field of video tracking of fast moving targets, can solve the problems of not meeting the real-time tracking requirements, high tracking time consumption, lack of template update, etc., to improve tracking accuracy and reduce error tracking rate , the effect of enhancing the classification performance

Active Publication Date: 2020-08-11
ZHEJIANG UNIV OF TECH
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In order to realize the tracking of fast moving objects, literature [5] proposed a fast target tracking method combining Kalman filter and Meanshift. Under the Meanshift framework, the Kalman filter is used to correct and predict the position of fast moving targets. This method is better than Camshift The calculation time of the method is less, but only color information is used in feature description, and the necessary template update is lacking, so it is not suitable for tracking in complex situations
At the same time, the tracking methods of these fast-moving targets all have a shortcoming that the tracking time consumes a lot and does not meet the real-time tracking requirements.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Fast Moving Target Tracking Method Based on Kalman Target Prediction and Multi-feature Compression Fusion
  • Fast Moving Target Tracking Method Based on Kalman Target Prediction and Multi-feature Compression Fusion
  • Fast Moving Target Tracking Method Based on Kalman Target Prediction and Multi-feature Compression Fusion

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0073] 1. Example 1: Tracking a single fast moving target

[0074] As shown in Figure 1, there is only one unobstructed fast-moving target in the Diving video sequence, and the target rotates and tilts. Figure 1 (a)-Figure 1 (c) shows the tracking effect of the CT method, Kalman+Meanshift and the method of the present invention on a fast moving target. From the experimental results of the five selected frames from frame 99 to frame 176, during this period, the athletes are completing two 360-degree rotations in the air, almost maintaining a uniform posture for rotation. Both methods can track athletes, but due to the high speed, the tracking frames of the other two algorithms have a certain degree of deviation. After the 187th frame, the athlete completed the air-turning motion, quickly changed the posture and quickly entered the water. At this time, the original CT method lost the target. Although the Kalman+Meanshift method can track the target, it also has a large deviation. ...

Embodiment 2

[0075] 2. Embodiment 2: Fast moving target tracking interfered by multiple similar objects

[0076] From the video sequence of Figure 2(a)-Figure 2(c), it can be seen that from the 110th frame, the wild goose begins to enter the cloud layer and is blocked by the cloud until it is completely blurred. The existing CT method loses the tracking target when the wild goose enters the cloud. The Kalman+Meanshift method turns the tracking target to another similar wild goose after the wild goose enters the cloud, and the tracking error occurs. However, the present invention can obtain a good tracking effect. When the target enters the cloud layer until it is gradually blurred, the lock is kept not lost.

Embodiment 3

[0077] 3. Embodiment 3: Fast moving target tracking with occlusion of similar objects

[0078] Figure 3 (a)-Figure 3 (c) also appeared in the interference of similar objects to the target, and at the same time, it was blocked by two players from the 21st frame. The other two algorithms both pointed the tracking object to the other outermost white clothing. Players. And because the high-resolution fusion color feature is introduced in this algorithm, it can be better distinguished from other players, even if it is blocked by a similar target, interference can be eliminated; at the same time, due to the role of predicting the target, the target is blocked from the It can achieve accurate positioning of the target when it comes out.

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention provides a fast moving target tracking method based on the Kalman target prediction multi-feature compression fusion. The fast moving target tracking method comprises a first step of compressing multi-features in a feature extraction stage to reduce the feature dimensions and shorten training time, a second time of conducting self-adaptive fusion on the compressed multi-features by using background weighted technology, forming feature vector for tracking to effectively reduce the interference of similar objects and the shielding object, and a third step of predicting the target position by using Kalman filtering in a tracking process and conducting target tracking optimization on aspects including 1) in a classification process, calculating the distance weight between the sample position and the kalman prediction position, inputting the weight in a bayesian classifier to enhance the classification performance and reduce error tracking rate; and (2) in a parameter updatingprocess, adopting a self-adaptive learning rate to replace a constant learning rate, so as to reduce interference of noise to the classification performance. According to the method, the tracking accuracy of a fast moving object under the complex condition can be effectively improved.

Description

[0001] (1) Technical field [0002] The invention relates to a video tracking method for fast moving targets. [0003] (2) Background technology [0004] The traditional compressed sensing method has a good tracking effect for slow-moving objects. However, once the tracking drift occurs in the context of the target's sustained and fast-moving, it will cause the accumulation of errors to find it difficult to retrieve the target. At present, most researches on fast target tracking are based on Kalman filter. [0005] Kalman filter is a method of estimating the linear minimum mean square error of the system state through recursive filtering. This method can get a good estimate of the target's speed, position, etc., so it has been widely used in target tracking. Applying this idea to the tracking of fast moving targets, Kalman filtering can predict the target from different angles and correct the target position in time, which can greatly improve the accuracy of tracking. In reference to...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06T7/277G06T7/246
Inventor 张霓章承成何熊熊
Owner ZHEJIANG UNIV OF TECH