Embodiments of this software relate to the sector of computer visual applied sciences, and specifically, to a goal tracking methodology and apparatus, a computer storage medium, and a machine. Target tracking is likely one of the hotspots in the sphere of computer imaginative and prescient research. Target tracking is widely utilized in a plurality of fields equivalent to video surveillance, ItagPro navigation, military, portable tracking tag human-laptop interaction, digital actuality, and autonomous driving. Simply put, goal tracking is to analyze and observe a given goal in a video to find out a precise location of the target in the video. Embodiments of this software present a target monitoring methodology and travel security tracker apparatus, a medium, and itagpro tracker a device, to effectively forestall prevalence of instances reminiscent of losing a tracking goal and itagpro tracker a monitoring drift, to make sure the accuracy of target monitoring. FIG. 1 is a schematic diagram of an application scenario of a goal monitoring methodology in keeping with an embodiment of this software. FIG. 2 is a schematic flowchart of a target tracking method based on an embodiment of this utility.

FIG. 11 is a schematic structural diagram of another target monitoring apparatus based on an embodiment of the current application. FIG. 12 is a schematic structural diagram of a target tracking device in line with an embodiment of this application. FIG. Thirteen is a schematic structural diagram of one other goal tracking device in response to an embodiment of this software. Features are normally numeric, but structural options akin to strings and graphs are utilized in syntactic pattern recognition. Web server. During precise software deployment, the server could also be an unbiased server, or a cluster server. The server could concurrently provide goal tracking providers for a plurality of terminal units. FIG. 1 is a schematic diagram of an utility state of affairs of a goal monitoring technique in accordance with an embodiment of this software. A hundred and one and a server 102 . One hundred and one is configured to send a video stream recorded by the surveillance digital camera 101 to the server 102 .

102 is configured to carry out the target tracking technique provided in this embodiment of this application, to perform goal tracking in video frames included in the video stream sent by the surveillance camera one zero one . 102 retrieves the video stream shot by the surveillance digicam 101 , and performs the following information processing for each video frame in the video stream: the server 102 first performs detection in an total vary of a current video frame by using 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 regions in the present video frame through the use of a function extraction mannequin, and calculates a feature similarity corresponding to every candidate area based on the deep feature corresponding to the every candidate area and a deep feature of the goal detected in a previous video body; and the server 102 further determines, in accordance with the characteristic similarity corresponding to the each candidate area, iTagPro tracker the target detected in the earlier video frame.

102 first performs target detection in the general range of the present video body by using the goal detection mannequin, to determine all the candidate regions existing in the present video frame, iTagPro tracker after which performs target monitoring based mostly on all the determined candidate areas, thereby enlarging a target monitoring range in every video body, so that occurrence of a case of losing a monitoring goal attributable to excessively quick motion of the monitoring goal can be successfully prevented. 102 additionally extracts the deep options of the candidate areas through the use of the feature extraction model, and determines the monitoring goal in the current video frame primarily based on the deep features of the candidate regions and the deep characteristic of the target detected within the earlier video body. Therefore, performing target tracking based on the deep feature can make sure that the decided tracking goal is more correct, and successfully prevent a case of a tracking drift.

FIG. 1 is only an instance. FIG. 2 is a schematic flowchart of a target tracking technique in line with an embodiment of this software. It is to be understood that the execution body of the target monitoring methodology just isn't limited only to a server, but additionally could also be applied to a system having a picture processing function reminiscent of a terminal machine. When the server needs to perform target monitoring for a primary video stream, the server obtains the first video stream, iTagPro tracker and performs a knowledge processing procedure shown in FIG. 2 for iTagPro shop a video body in the first video stream, iTagPro technology to trace a target in the primary video stream. Further, the information processing procedure proven in FIG. 2 is performed for a video body in the obtained first video stream, to implement goal tracking in the first video stream. FIG. 2 for a video body in the first video stream, iTagPro tracker to implement target monitoring in the primary video stream.