For example, the person extraction unit 10 creates a pair of the detection ID obtained from the frame to be processed and the detection ID obtained from the previously processed frame, and calculates a similarity between the feature values associated with the respective detection IDs for each pair. Then, the person extraction unit 10 determines whether the similarity is equal to or higher than the predetermined level.

In a case where a person whose similarity is equal to or higher than the predetermined level has not been detected in the previously processed frame, the person extraction unit 10 assigns a person ID in association with the detection ID of the person, and newly registers the person ID in the detected person information shown in FIG. 6.

On the other hand, in a case where a person whose similarity is equal to or higher than the predetermined level has been detected in the previously processed frame, the person extraction unit 10 registers the detection ID of the person in the detected person information shown in FIG. 6 so as to be associated with the same person ID as the detection ID determined to be similar.

Thereafter, the person extraction unit 10 performs similar processing for all frames to be detected.

Here, a “specific example of processing for determining whether or not the outer appearance feature value of each person detected from the frame to be processed is similar by the predetermined level or more to that of any of persons detected in previously processed frames” will be described.