Software-based object tracking method and computing device therefor

A software-based object tracking method that is a method of providing object tracking performed in a computing device including a camera module fixed in a preset forward direction to generate a captured frame image, the software-based object tracking method comprising:

• receiving a plurality of captured frame images time sequentially captured at a first resolution from the camera module;
• selecting at least one selected frame image by extracting at least some of the plurality of captured frame images;
• identifying whether a tracking object exists in the at least one selected frame image;
• identifying the tracking object in the at least one selected frame image using a first deep learning model;
• generating feature data for the tracking object using a second deep learning model;
• determining whether the tracking object identified in different selected frame images corresponds to a same object based on a similarity between feature data;
• setting each of partial areas of the at least one selected frame image including the tracking object as the viewing window on the basis of a location of the tracking object within the at least one selected frame image;
• controlling the viewing window to automatically move and be size-adjusted in response to the location and scale of the tracking object;
• processing the at least one selected frame image within the viewing window in a manner corresponding to panning or zooming on the tracking object while the camera module remains fixed;
• determining positional criticality by calculating a distance between a first viewing window set for a first selected frame image and a second viewing window set for a second selected frame image;
• and
• resetting the second viewing window based on the first viewing window when the positional criticality is not satisfied.

A computing device comprising:

• a memory configured to store one or more instructions;
• and
• at least one processor configured to execute the one or more instructions stored in the memory,
• wherein the at least one processor executes the one or more instructions to receive a selected frame image captured at a first resolution from the camera module,
• set a second resolution for a viewing window,
• identify whether a tracking object exists in the selected frame image, and
• set a partial area of the selected frame image including the tracking object as the viewing window on the basis of a location of the tracking object within the selected frame image, and
• the second resolution of the viewing window is a resolution lower than the first resolution of the selected frame image,
• wherein the at least one processor is further configured to identify the tracking object using a first deep learning model, and to determine whether the tracking object identified in different selected frame images corresponds to a same object by generating feature data using a second deep learning model and determining the same object based on a similarity between feature data,
• and wherein the at least one processor is further configured to determine positional criticality by calculating a distance between a first viewing window set for a first selected frame image and a second viewing window set for a second selected frame image, and to reset the second viewing window based on the first viewing window when the positional criticality is not satisfied,
• wherein the at least one processor is further configured to control the viewing window to automatically move and be size-adjusted in response to the location and scale of the tracking object, and to process the selected frame image within the viewing window in a manner corresponding to panning or zooming on the tracking object while the camera module remains fixed.