US20210019627A1 - Target Tracking Method and Apparatus, Medium, And Device - Google Patents > 자유게시판

• 목록
• 답변
• 글쓰기
• 게시판 리스트 옵션
  • 수정
  • 삭제

Embodiments of this application relate to the sector of pc visual technologies, iTagPro features and in particular, to a target monitoring technique and apparatus, a pc storage medium, and a system. Target tracking is one of the hotspots in the sector of laptop vision analysis. Target tracking is widely utilized in a plurality of fields reminiscent of video surveillance, navigation, military, human-laptop interaction, virtual reality, and autonomous driving. Simply put, target monitoring is to investigate and track a given target in a video to find out an actual location of the goal within the video. Embodiments of this utility present a goal tracking method and apparatus, a medium, iTagPro features and a machine, iTagPro features to effectively prevent prevalence of instances equivalent to losing a monitoring goal and a tracking drift, to make sure the accuracy of target monitoring. FIG. 1 is a schematic diagram of an utility state of affairs of a target tracking method based on an embodiment of this application. FIG. 2 is a schematic flowchart of a target tracking method in line with an embodiment of this utility.

FIG. Eleven is a schematic structural diagram of one other target monitoring apparatus in keeping with an embodiment of the current utility. FIG. 12 is a schematic structural diagram of a goal tracking device based on an embodiment of this application. FIG. 13 is a schematic structural diagram of another goal tracking device based on an embodiment of this utility. Features are usually numeric, but structural features similar to strings and graphs are used in syntactic pattern recognition. Web server. During precise application deployment, the server may be an independent server, or a cluster server. The server may simultaneously provide goal tracking providers for iTagPro shop a plurality of terminal units. FIG. 1 is a schematic diagram of an software scenario of a goal tracking methodology in response to an embodiment of this software. One hundred and one and a server 102 . A hundred and one is configured to send a video stream recorded by the surveillance digicam a hundred and one to the server 102 .

102 is configured to perform the goal monitoring methodology provided in this embodiment of this utility, to perform goal tracking in video frames included within the video stream despatched by the surveillance digital camera 101 . 102 retrieves the video stream shot by the surveillance digital camera one zero one , and performs the next information processing for iTagPro features each video body within the video stream: the server 102 first performs detection in an total vary of a current video body by utilizing a target detection mannequin, to obtain all candidate regions existing in the video body; the server 102 then extracts deep features respectively corresponding to all of the candidate areas in the present video body through the use of a feature extraction model, ItagPro and calculates a characteristic similarity corresponding to every candidate area in response to the deep characteristic corresponding to the every candidate region and a deep characteristic of the target detected in a earlier video frame; and the server 102 additional determines, based on the characteristic similarity corresponding to the each candidate region, iTagPro features the goal detected within the previous video body.

102 first performs goal detection in the overall vary of the current video frame through the use of the goal detection model, to determine all the candidate areas current in the current video frame, after which performs goal monitoring based mostly on all of the determined candidate areas, thereby enlarging a target tracking range in every video body, in order that incidence of a case of losing a monitoring goal because of excessively fast movement of the tracking target can be successfully prevented. 102 also extracts the deep iTagPro features of the candidate areas through the use of the feature extraction mannequin, and determines the tracking target in the present video frame primarily based on the deep features of the candidate areas and the deep feature of the target detected within the previous video body. Therefore, performing target monitoring primarily based on the deep feature can be certain that the decided monitoring target is more accurate, and successfully prevent a case of a tracking drift.

FIG. 1 is barely an example. FIG. 2 is a schematic flowchart of a target tracking method in keeping with an embodiment of this utility. It is to be understood that the execution physique of the target tracking methodology shouldn't be limited solely to a server, but additionally could also be utilized to a system having a picture processing perform equivalent to a terminal system. When the server needs to carry out goal tracking for a first video stream, iTagPro bluetooth tracker the server obtains the first video stream, and performs an information processing process proven in FIG. 2 for a video frame in the primary video stream, to track a target in the first video stream. Further, the information processing procedure proven in FIG. 2 is carried out for a video frame within the obtained first video stream, to implement goal tracking in the primary video stream. FIG. 2 for a video frame in the first video stream, to implement target monitoring in the primary video stream.