Video searching

A processor configured to:

• a. receive a target selected from within video data collected from a first camera view, the target defined by a descriptor vector;
• b. identify video data collected from a set of second camera views imaging one or more locations different to a location imaged in the first camera view, the video data identified according to space-time profiling based on the distance between the location imaged in the first camera view and the location imaged in each of the set of second camera views and the difference between the time of collection of the first camera view compared to the time of collection of each of the set of second camera views, the identified video data comprising the set of second camera views wherein a probability of identifying a candidate matching the target according to the space-time profiling is greater than a given threshold;
• c. identify a selection of candidate matches to the target within the identified video data, each candidate match defined by a descriptor vector;
• d. rank the candidate matches wherein the ranking is a relative ranking comprising a ranking of the candidate matches according to the relevance of different visual features within the appearance descriptor when comparing each candidate match to the selected target, and a ranking of the candidate matches according to overall match confidence;
• e. receive a signal confirming or rejecting one or more of the ranked candidate matches as a match to the target;
• f. identify a further selection of candidate matches from further video data by repeating steps b to d wherein each of the confirmed candidate matches at step e is used as a target;
• and
• g. indicate portions of the further video data containing the further selection of candidate matches to the target.

One or more non-transitory computer readable media storing computer readable instructions that, when executed, cause a system to:

• a. receive a target selected from within video data collected from a first camera view, the target defined by a descriptor vector;
• b. identify video data collected from a set of second camera views imaging one or more locations different to a location imaged in the first camera view, the video data identified according to space-time profiling based on the distance between the location imaged in the first camera view and the location imaged in each of the set of second camera views and the difference between the time of collection of the first camera view compared to the time of collection of each of the set of second camera views, the identified video data comprising the set of second camera views wherein a probability of identifying a candidate matching the target according to the space-time profiling is greater than a given threshold
• c. identify a selection of candidate matches to the target within the identified video data, each candidate match defined by a descriptor vector;
• d. rank the candidate matches wherein the ranking is a relative ranking comprising a ranking of the candidate matches according to the relevance of different visual features within the appearance descriptor when comparing each candidate match to the selected target, and a ranking of the candidate matches according to overall match confidence;
• e. receive a signal confirming or rejecting one or more of the ranked candidate matches as a match to the target;
• f. identify a further selection of candidate matches from further video data by repeating steps b to d wherein each of the confirmed candidate matches at step e is used as a target;
• and
• g. indicate portions of the further video data containing the further selection of candidate matches to the target.