Object trajectory proposing method based on hierarchical spatial-temporal region merging
A hierarchical and regional technology, applied in the field of computer vision, can solve the problems of poor object trajectory proposal effect and failure to consider the characteristics of video changes, etc., achieve high average trajectory intersection ratio and recall rate, and improve the effect
Active Publication Date: 2018-11-27
NANJING UNIV
View PDF1 Cites 4 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0005] The problem to be solved by the present invention is: the existing object trajectory proposal method mainly relies on the image proposal, and then derives the video from the image, without considering the changing characteristics of the video, and failing to use the spatiotemporal characteristics of the entire video to estimate the object-likeness, resulting in Object trajectory proposals do not perform well
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
[0037] The present invention proposes an object trajectory proposal method based on hierarchical spatio-temporal region merging, such as figure 1 shown, including the following steps:
[0038] 1) Construct a hierarchical spatio-temporal region representation of the video by matching the hierarchical video frame regions of adjacent frames:
[0039] 1.1) Using the ultra-segmentation contour map (ultrametric contour map) method on each video frame f, the video frame is divided into multiple basic video frame areas, as the leaf node area in the hierarchical video frame area representation, the leaf node area The set formed is L f , where the video frame area refers to the segmented area on the video frame, and the hyper-segmentation contour map method results in a small area, that is, the leaf node area, see reference [2] for details. Iteratively merge adjacent and similar video frame areas, where the video frame area includes the leaf node area and the merged video frame area, e...
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
Provided is an object trajectory proposing method based on hierarchical spatial-temporal region merging. The method comprises: constructing a hierarchical spatial-temporal region representation of a video by matching a hierarchical video frame region of an adjacent frame on a video, and then generating candidate object trajectories by merging the hierarchical spatial-temporal regions, finally, using a comprehensive scoring mechanism that combines appearance objectness and movement objectness to score the candidate object trajectories, and using the candidate object trajectories with the highest scores as a final object trajectory proposal result. The method combines appearance information and movement information of an object in the video, and time-space characteristics of the whole videoare used to produce the object trajectory proposal result, and better effect than that of an existing method is achieved.
Description
technical field [0001] The invention belongs to the technical field of computer vision, and relates to an object trajectory proposal method for video, specifically a method for object trajectory proposal based on hierarchical spatio-temporal region merging Background technique [0002] The goal of object trajectory proposal is to identify regions in a video that may contain objects with a limited number of sequences of bounding boxes. As a basic technology in the field of computer vision, this technology can be widely used in video-oriented object recognition, motion recognition, object segmentation, content retrieval and other fields. It is an important and novel research direction. [0003] The basic strategy of the existing object trajectory proposal method is: use the image object proposal method on one or several frames of the video to obtain the bounding box that may contain objects on a single video frame, and then use the object tracker to track the object on the rem...
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): G06T7/215G06T7/246
CPCG06T7/215G06T7/246G06T2207/10016
Inventor 任桐炜武港山孙旭王元天
Owner NANJING UNIV