First, the person extraction unit 10 sets a plurality of detection IDs located in the first layer as detection IDs to be compared. The person extraction unit 10 creates a pair of the detection ID of a person detected from the frame to be processed and each of the plurality of detection IDs located in the first layer. Then, the person extraction unit 10 calculates a similarity for each pair, and determines whether the calculated similarity is equal to or higher than the first threshold value.

In a case where there is no detection ID whose similarity is equal to or higher than the first threshold value in the first layer, the person extraction unit 10 determines that the outer appearance feature value of the person detected from the frame to be processed is not similar by the predetermined level or more to that of any of persons detected in previously processed frames. Then, the person extraction unit 10 associates a new person ID with the detection ID of the person detected from the frame to be processed, and registers the new person ID in the detected person information shown in FIG. 6.

On the other hand, in a case where there is a detection ID whose similarity is equal to or higher than the first threshold value in the first layer, the person extraction unit 10 moves the comparison subject to the second layer. Specifically, the group in the second layer linked to the “detection ID of the first layer determined that the similarity is equal to or higher than the first threshold value” is set as a comparison subject.