Multi-scale target sensing tracking method based on twin network
A twin network, multi-scale technology, applied in the field of image processing, can solve the problem of difficult matching of related features
Pending Publication Date: 2021-05-04
ZHEJIANG UNIV OF TECH
View PDF0 Cites 6 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Afterwards, a fixed size (square) receptive field will make it difficult to match relevant features to objects of different shapes
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0060] The present invention will be further described below in conjunction with the accompanying drawings.
[0061] refer to Figure 1 to Figure 7 , a multi-scale object-aware tracking method based on Siamese networks, including the following steps:
[0062] S1. Fine feature aggregation, the process is as follows:
[0063] S1.1 The picture I obtained according to the first frame of the video sequence 1 and the target's bounding box information B 1 , slice and deform to obtain the tracking template Z 1 ,Such as figure 2 shown; in the follow-up tracking process, according to the tracking results of the previous frame B i-1 , for the input picture I i , i∈[2,n] slices and deforms to obtain the search image X i ,Such as image 3 shown;
[0064] S1.2 The tracking template Z to be obtained 1 and search image X i Input the pre-trained deep residual network "ResNet-50" to obtain deep features with The model will collect the features output by the last three layers (CO...
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
Login to View More Abstract
The invention discloses a multi-scale target sensing tracking method based on a twin network, and the method comprises the steps: cutting a block from a picture as a template picture in a first frame according to a marked target bounding box, inputting the template picture into a ResNet network, and extracting the features of the template picture; in a subsequent tracking process, firstly, a search area picture with a certain size being cut in a current frame according to target position information of a previous frame; secondly, inputting the same ResNet network to extract features of the ResNet network, and performing pixel-level correlation calculation on the features of the ResNet network and template features to obtain correlation features with similarity information of the ResNet network and the template features; enabling the related features to pass through a channel attention module, a non-local attention module and an anchor-free prediction module in sequence, and obtaining a classification graph and a consistent frame regression result; and finally, finding out the position with the highest probability of the positive sample in the classification graph, and finding out a corresponding object frame according to the position. And after the positioning information of the current frame target is predicted, a next frame cycle is executed.
Description
technical field [0001] The invention belongs to the field of image processing, and relates to a twin network-based multi-scale target perception tracking method. Background technique [0002] Target tracking is one of the important research directions in the field of computer vision, and it has a wide range of applications in military and civilian fields such as robotics, human-computer interaction, military reconnaissance, intelligent transportation, and virtual reality. In recent years, many scholars have done a lot of work on object tracking and made some progress. However, in complex environments, there are still problems such as target appearance deformation (target texture, shape, pose changes, etc.), illumination changes, fast motion and motion blur, similar background interference, plane rotation inside and outside, scale changes, occlusion and out of view, etc. Stable real-time object tracking in complex environments remains a challenging problem. [0003] The Sia...
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
Login to View More IPC IPC(8): G06T7/246G06K9/62G06N3/04G06N3/08
CPCG06T7/246G06N3/08G06T2207/10016G06T2207/20081G06T2207/20084G06N3/047G06N3/048G06F18/2415G06F18/241
Inventor 产思贤陶健周小龙白琮郏杨威郑竟成陈胜勇
Owner ZHEJIANG UNIV OF TECH