Joint detection and tracking method, device and equipment based on multi-dimensional attention mechanism
A joint detection and attention technology, applied in the field of multi-target tracking, can solve problems such as feature decoupling and achieve the effect of enhancing robustness
Pending Publication Date: 2022-06-24
TSINGHUA UNIV
View PDF0 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0005] This application provides a joint detection and tracking method, device and equipment based on a multi-dimensional attention mechanism to solve the problems of cross-temporal and spatial extraction of targets of different scales in joint detection and tracking tasks and feature decoupling in different tasks.
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
[0057] Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary, and are intended to be used to explain the present application, but should not be construed as a limitation to the present application.
[0058] The method, apparatus, and device for joint detection and tracking based on a multi-dimensional attention mechanism according to embodiments of the present application are described below with reference to the accompanying drawings. Most of the existing joint detection and tracking methods mentioned in the above-mentioned background technology center overly rely on the appearance features of the target, the target is easily occluded in complex traffic scenes, and t...
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 relates to the technical field of multi-target tracking, in particular to a joint detection and tracking method, device and equipment based on a multi-dimensional attention mechanism, and the method comprises the steps: carrying out the feature extraction and fusion processing of a current image of a current frame at a current moment, and generating an original feature map; based on the original feature map, generating a decoupling feature map, executing a preset detection task and a tracking task by using decoupling features, and obtaining a detection result and tracking target features of the current frame; and constructing a multi-feature cascade tracker by taking the detection result of the current frame, the tracking target feature and the target feature in the historical trajectory as input, and generating a final detection and tracking result. Therefore, the cross-space-time extraction problem of targets of different scales in the joint detection and tracking task and the feature decoupling problem in different tasks are solved, and the robustness of a tracking algorithm in a complex scene is enhanced by constructing a multi-feature cascade tracker.
Description
technical field [0001] The present application relates to the technical field of multi-target tracking, and in particular, to a joint detection and tracking method, apparatus and device based on a multi-dimensional attention mechanism. Background technique [0002] Different from the classical first detection and tracking implementation in the multi-target tracking field, the joint detection and tracking method extracts the features of target detection and tracking from a single neural network, which improves the running efficiency of the model and reduces the computational cost. For the joint detection and tracking method, the current technical implementation schemes and typical algorithms can be divided into the following two types. [0003] (1) Based on the positional correlation of objects in the front and rear frames. Trackor, as a representative joint detection and tracking method, uses the bounding box detection result of the previous frame as the candidate region of...
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 Patent Type & Authority Applications(China)
IPC IPC(8): G06V20/40G06K9/62G06V10/80G06V10/75
CPCG06F18/22G06F18/25
Inventor 郑四发张创吴浩然许庆王建强李克强
Owner TSINGHUA UNIV