Visual tracking in video images in unconstrained environments by exploiting on-the-fly context using supporters and distracters

A method of comprising:

• receiving video data;
• tracking a single target identified in the video data, the tracking comprising detecting and using one or more distracters, which have an appearance similar to that of the tracked target in that each of the one or more distracters is a likely false positive for a classifier that detects the single target, wherein the tracking comprises detecting and using one or more supporters, which include points outside of the tracked target that have motion correlation with the tracked target;
• limiting a maximum number of the one or more distracters and a maximum number of the one or more supporters;
• and
• outputting information corresponding to the tracked target.

A system comprising:

• a video acquisition system;
• and
• processor electronics configured to track a single target identified in video data from the video acquisition system, including detecting and using one or more distracters, which have an appearance similar to that of the tracked target in that each of the one or more distracters is a likely false positive for a classifier that detects the single target, and to output information corresponding to the tracked target;
• wherein the processor electronics are configured to detect and use one or more supporters, which include points outside of the tracked target that have motion correlation with the tracked target, to track the single target;
• and
• wherein the processor electronics are configured to detect the target and the one or more distracters using a shared classifier having binary testing functions corresponding to positive and negative samples added during training, but not including hard negative samples.

A system comprising:

• a video acquisition system;
• and
• processor electronics configured to track a single target identified in video data from the video acquisition system, including detecting and using one or more distracters, which have an appearance similar to that of the tracked target in that each of the one or more distracters is a likely false positive for a classifier that detects the single target, and to output information corresponding to the tracked target;
• wherein the processor electronics are configured to detect and use one or more supporters, which include points outside of the tracked target that have motion correlation with the tracked target, to track the single target;
• and
• wherein the processor electronics are configured to limit a maximum number of the one or more distracters and a maximum number of the one or more supporters.