Recover keypoint-based target tracking from occlusion using deep neural network segmentation

1. A method, implemented by an information handling system comprising a processor and a memory accessible by the processor, the method comprising:

capturing a plurality of digital images of an area, wherein the area includes a selected moving object and wherein the plurality of digital images are taken in a sequential time-based order;

determining that a keypoint-based (KP-based) matching model used to track the selected moving object by analyzing the plurality of digital images is no longer tracking the selected moving object, wherein the determining is based on comparing a neural network based (NN-based) location of the selected moving object with a KP-based location of the selected moving object;

identifying the neural network based (NN-based) location of the selected moving object in a selected one of the plurality of digital images using a NN-based matching model;

providing the NN-based location and the selected digital image to the KP-based matching model; and

re-establishing tracking of the selected moving object in one or more digital images subsequent to the selected digital image using the KP-based matching model.

2. The method of claim 1 wherein the determining that the KP-based matching model is no longer tracking the selected moving object further comprises:

identifying a KP-based location in the selected digital image corresponding to where the KP-based matching model locates the selected moving object; and

comparing the KP-based location to the NN-based location, wherein the determination is based on the comparison.

3. The method of claim 1 further comprising:

failing to identify a KP-based location in a set of one or more of the plurality of digital images due to an occlusion of the selected moving object; and

identifying the NN-based location in the selected digital image that is captured after the capturing of at least one of the set of the digital images that was captured while the selected moving object was occluded.

4. The method of claim 1 further comprising:

initiating the tracking of the selected moving object by capturing a plurality of features pertaining to the selected moving object by the NN-based matching model; and

providing a set of the features to the KP-based matching model to commence tracking of the selected moving object by the KP-based tracking model.

5. The method of claim 1 further comprising:

tracking the selected moving object in near-real-time by the KP-based matching model by locating the selected moving object in most of the plurality of digital images; and

tracking the selected moving object in a fraction of the plurality of digital images by the NN-based matching model, wherein the KP-based matching model has a faster execution time than the NN-based matching model.

6. The method of claim 5 further comprising:

comparing the NN-based location identified during the tracking by the NN-based matching model with the KP-based location identified by the KP-based matching model, wherein the NN-based location and the KP-based location are identified in a common digital image; and

determining that the KP-based matching model is no longer tracking the selected moving object in response to the KP-based location differing from the NN-based location.

7. The method of claim 1 further comprising:

training the NN-based matching model to identify the selected moving object prior to the tracking of the selected moving object.

8. An information handling system comprising:

one or more processors;

a memory coupled to at least one of the processors; and

a set of computer program instructions stored in the memory and executed by at least one of the processors in order to perform actions comprising:

capturing a plurality of digital images of an area, wherein the area includes a selected moving object and wherein the plurality of digital images are taken in a sequential time-based order;

determining that a keypoint-based (KP-based) matching model used to track the selected moving object by analyzing the plurality of digital images is no longer tracking the selected moving object, wherein the determining is based on comparing a neural network based (NN-based) location of the selected moving object with a KP-based location of the selected moving object;

identifying the neural network based (NN-based) location of the selected moving object in a selected one of the plurality of digital images using a NN-based matching model;

providing the NN-based location and the selected digital image to the KP-based matching model; and

re-establishing tracking of the selected moving object in one or more digital images subsequent to the selected digital image using the KP-based matching model.

9. The information handling system of claim 8 wherein the determining that the KP-based matching model is no longer tracking the selected moving object further comprises:

identifying a KP-based location in the selected digital image corresponding to where the KP-based matching model locates the selected moving object; and

comparing the KP-based location to the NN-based location, wherein the determination is based on the comparison.

10. The information handling system of claim 8 wherein the actions further comprise:

failing to identify a KP-based location in a set of one or more of the plurality of digital images due to an occlusion of the selected moving object; and

identifying the NN-based location in the selected digital image that is captured after the capturing of at least one of the set of the digital images that was captured while the selected moving object was occluded.

11. The information handling system of claim 8 wherein the actions further comprise:

initiating the tracking of the selected moving object by capturing a plurality of features pertaining to the selected moving object by the NN-based matching model; and

providing a set of the features to the KP-based matching model to commence tracking of the selected moving object by the KP-based tracking model.

12. The information handling system of claim 8 wherein the actions further comprise:

tracking the selected moving object in near-real-time by the KP-based matching model by locating the selected moving object in most of the plurality of digital images; and

tracking the selected moving object in a fraction of the plurality of digital images by the NN-based matching model, wherein the KP-based matching model has a faster execution time than the NN-based matching model.

13. The information handling system of claim 12 wherein the actions further comprise:

comparing the NN-based location identified during the tracking by the NN-based matching model with the KP-based location identified by the KP-based matching model, wherein the NN-based location and the KP-based location are identified in a common digital image; and

determining that the KP-based matching model is no longer tracking the selected moving object in response to the KP-based location differing from the NN-based location.

14. The information handling system of claim 8 wherein the actions further comprise:

training the NN-based matching model to identify the selected moving object prior to the tracking of the selected moving object.

15. A computer program product comprising:

a computer readable storage medium, comprising computer program code that, when executed by an information handling system, executes a digital assistant that performs actions comprising:

capturing a plurality of digital images of an area, wherein the area includes a selected moving object and wherein the plurality of digital images are taken in a sequential time-based order;

determining that a keypoint-based (KP-based) matching model used to track the selected moving object by analyzing the plurality of digital images is no longer tracking the selected moving object, wherein the determining is based on comparing a neural network based (NN-based) location of the selected moving object with a KP-based location of the selected moving object;

identifying the neural network based (NN-based) location of the selected moving object in a selected one of the plurality of digital images using a NN-based matching model;

providing the NN-based location and the selected digital image to the KP-based matching model; and

re-establishing tracking of the selected moving object in one or more digital images subsequent to the selected digital image using the KP-based matching model.

16. The computer program product of claim 15 wherein the determining that the KP-based matching model is no longer tracking the selected moving object further comprises:

identifying a KP-based location in the selected digital image corresponding to where the KP-based matching model locates the selected moving object; and

comparing the KP-based location to the NN-based location, wherein the determination is based on the comparison.

17. The computer program product of claim 15 wherein the actions further comprise:

failing to identify a KP-based location in a set of one or more of the plurality of digital images due to an occlusion of the selected moving object; and

identifying the NN-based location in the selected digital image that is captured after the capturing of at least one of the set of the digital images that was captured while the selected moving object was occluded.

18. The computer program product of claim 15 wherein the actions further comprise:

initiating the tracking of the selected moving object by capturing a plurality of features pertaining to the selected moving object by the NN-based matching model; and

providing a set of the features to the KP-based matching model to commence tracking of the selected moving object by the KP-based tracking model.

19. The computer program product of claim 15 wherein the actions further comprise:

tracking the selected moving object in near-real-time by the KP-based matching model by locating the selected moving object in most of the plurality of digital images; and

tracking the selected moving object in a fraction of the plurality of digital images by the NN-based matching model, wherein the KP-based matching model has a faster execution time than the NN-based matching model.

20. The computer program product of claim 19 wherein the actions further comprise:

comparing the NN-based location identified during the tracking by the NN-based matching model with the KP-based location identified by the KP-based matching model, wherein the NN-based location and the KP-based location are identified in a common digital image; and

determining that the KP-based matching model is no longer tracking the selected moving object in response to the KP-based location differing from the NN-based location.