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

End-to-end target tracking method based on hierarchical feature representation

A target tracking and feature extraction technology, applied in the field of deep neural network, can solve the problems of large amount of calculation and time-consuming, and achieve the effect of improving the effect

Active Publication Date: 2018-11-16
HANGZHOU DIANZI UNIV
View PDF7 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0008] (2) Processing video is usually computationally intensive and relatively time-consuming

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
  • End-to-end target tracking method based on hierarchical feature representation
  • End-to-end target tracking method based on hierarchical feature representation
  • End-to-end target tracking method based on hierarchical feature representation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0045] The detailed parameters of the present invention will be further specifically described below.

[0046] Such as figure 1 As shown, the present invention provides a deep neural network framework for target tracking.

[0047] Step (1), data preprocessing, feature extraction

[0048] For the image pair (x', y'), where x' is the template image frame, the template image frame x' is preprocessed and scaled to a size of 127*127; y' is the search image frame, and the search image frame y' is processed The preprocessing is scaled to a size of 255*255; then a network flow of the Siamese network is used to calculate their respective feature representations. Here we use ImageNet's video target detection data set as training data and OTB-100 as test data. For image pair data, the existing Alexnet network model is used to extract image features. Specifically, the template image in the image data pair is scaled to 127×127, the search image size is scaled to 255×255, and input into...

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 discloses an end-to-end target tracking method based on hierarchical feature representation. The method comprises the following steps of (1) processing video frame data to obtain an input image pair and extracting features, (2) carrying out cross-correlation operation on a layered feature representation of the image pair and image pair features, (3) balancing a morphological featureand a semantic feature to obtain a final response graph through a mode of weight fusion for a response graph obtained by a layered cross-correlation operation, and (4) carrying out model training andtraining neural network parameters by using a back propagation algorithm. The invention provides a neural network model for target tracking and particularly provides a network structure based on full-volume integral layer feature fusion representation, and a relatively competitive effect in a target tracking field is obtained.

Description

technical field [0001] The present invention relates to a deep neural network for target tracking (Visual tracking), in particular to a method for uniformly modeling and expressing the layered features of an image and a modeling expression for template matching and fusion of the layered features of an image . Background technique [0002] Target tracking is an important research direction in computer vision and video analysis. Single target tracking aims to track a target based on a given target in the first frame of the video, and obtain the position of the target in subsequent frames of the video. [0003] With the rapid development of deep learning in recent years, end-to-end problem modeling using deep neural networks, such as deep convolutional neural networks (CNN), has become the mainstream of computer vision. research direction. In the single target tracking algorithm, the idea of ​​end-to-end modeling is introduced, the video frame itself is used as input, and the...

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
IPC IPC(8): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/084G06V20/54G06V20/40G06N3/045G06F18/25
Inventor 朱素果俞俊方振影
Owner HANGZHOU DIANZI UNIV