Analysis apparatus, analysis method, and storage medium
The analysis apparatus ( 2000 ) includes a co-appearance event extraction unit ( 2020 ) and a frequent event detection unit ( 2040 ). The co-appearance event extraction unit ( 2020 ) extracts co-appearance events of two or more persons from each of a plurality of sub video frame sequences. The sub video frame sequence is included in a video frame sequence. The analysis apparatus ( 2000 ) may obtain the plurality of sub video frame sequences from one or more of the video frame sequences. The one or more of the video frame sequences may be generated by one or more of surveillance cameras. Each of the sub video frame sequences has a predetermined time length. The frequent event detection unit ( 2040 ) detects co-appearance events of the same persons occurring at a frequency higher than or equal to a pre-determined frequency threshold.
1. An analysis apparatus comprising:
a co-appearance event extraction unit extracting a co-appearance event of two or more persons from each of a plurality of sub video frame sequences, the sub video frame sequence being included in a video frame sequence generated by a surveillance camera, each of the sub video frame sequences having a predetermined time length; and
a frequent event detection unit detecting co-appearance events of same persons occurring at a frequency higher than or equal to a pre-determined frequency threshold,
wherein in at least one co-appearance event, of the co-appearance events, of two or more persons, a first video frame from which a person is extracted is different from a second video frame from which another person is extracted,
wherein the at least one co-appearance event specifies each of the person and the another person, and
wherein the co-appearance event specifies that the unspecified combination of persons appeared in sequence in the video frame sequence among the predetermined time length and that the person appeared in the video frame sequence before the another person appeared in the video frame sequence.
2. The analysis apparatus according to claim 1 , further comprising an output unit outputting the co-appearance events of the same persons occurring at a frequency higher than or equal to the pre-determined frequency threshold, as an indication of a suspicious activity.
3. The analysis apparatus according to claim 1 ,
wherein the co-appearance event extraction unit assigns a label to each of feature-values extracted from video frames of the video frame sequences, the feature-value indicating a feature of a person, feature-values indicating features of the same person being assigned the same label, feature-values indicating features of different persons being assigned different labels, and
wherein the frequent event detection unit handles the co-appearance events of a first label and a second label as the co-appearance events of a first person and a second person, the first label being assigned to a feature-value indicating a feature of the first person, the second label being assigned to a feature-value indicating a feature of the second person.
4. The analysis apparatus according to claim 3 , wherein the co-appearance event extraction unit assigns labels to the feature-values by repeatedly performing for each of the feature-values:
determining whether or not there is a feature-value to which a label is already assigned and which is similar to a current feature-value;
when there is no feature-value to which a label is already assigned and which is similar to a current feature-value, assigning a new label to the current feature-value; and
when there is a feature-value to which a label is already assigned and which is similar to a current feature-value, assigning to the current feature-value the same label with a feature-value most similar to the current feature-value.
5. The analysis apparatus according to claim 3 , wherein the frequent event detection unit performs:
for each sub video frames, calculating a size-N (N is an integer greater than 0) combinations of labels assigned to the feature-values extracted from the sub video frame:
counting the number of each size-N combinations included in the sub video frames; and
detecting the size-N combinations of the same labels the number of which is greater than or equal to the pre-determined frequency threshold, as the co-appearance events of the same persons occurring at a frequency higher than or equal to the pre-determined frequency threshold.
6. An analysis method executed by a computer, the method comprising:
a co-appearance event extraction step of extracting a co-appearance event of two or more persons from each of a plurality of sub video frame sequences of one or more of video frame sequences generated by one or more of surveillance cameras, each of the sub video frame sequences having a predetermined time length; and
a frequent event detection step of detecting co-appearance events of same persons occurring at a frequency higher than or equal to a pre-determined frequency threshold,
wherein in at least one co-appearance event, of the co-appearance events, of two or more persons, a first video frame from which a person is extracted is different from a second video frame from which another person is extracted,
wherein the at least one co-appearance event specifies each of the person and the another person, and
wherein the co-appearance event specifies that the unspecified combination of persons appeared in sequence in the video frame sequence among the predetermined time length and that the person appeared in the video frame sequence before the another person appeared in the video frame sequence.
7. A non-transitory computer-readable storage medium storing a program causing a computer to execute each step of the analysis method according to claim 6 .
8. An analysis apparatus comprising:
a co-appearance event extraction unit extracting a co-appearance event of two or more persons from each of a plurality of sub video frame sequences, the sub video frame sequence being included in a video frame sequence generated by a surveillance camera, each of the sub video frame sequences having a predetermined time length; and
a frequent event detection unit detecting co-appearance events of same persons occurring at a frequency higher than or equal to a pre-determined frequency threshold,
wherein in at least one of the co-appearance events of two or more persons a first video frame from which a person is extracted is different from a second video frame from which another person is extracted,
wherein the co-appearance event extraction unit assigns a label to each of feature-values extracted from video frames of the video frame sequences, the feature-value indicating a feature of a person, feature-values indicating features of the same person being assigned the same label, feature-values indicating features of different persons being assigned different labels,
wherein the frequent event detection unit handles the co-appearance events of a first label and a second label as the co-appearance events of a first person and a second person, the first label being assigned to a feature-value indicating a feature of the first person, the second label being assigned to a feature-value indicating a feature of the second person, and
wherein the frequent event detection unit performs:
for each sub video frames, calculating a size-N (N is an integer greater than 0) combinations of labels assigned to the feature-values extracted from the sub video frame:
counting the number of each size-N combinations included in the sub video frames; and
detecting the size-N combinations of the same labels the number of which is greater than or equal to the pre-determined frequency threshold, as the co-appearance events of the same persons occurring at a frequency higher than or equal to the pre-determined frequency threshold.