Image monitoring system
An abandoned item detection unit detects an abandoned item shown in an image taken by any of a plurality of imaging devices. A person identification unit specifies the image directly before or directly after the appearance of the abandoned item as an image of abandonment timing, and identifies at least one person shown in the image of abandonment timing as a target person. A person search unit searches for an image showing the target person from among the images respectively taken by any of the plurality of imaging devices. A result display unit outputs a display showing the movement of the target person on the screen of a user terminal on the basis of the imaging device that has taken each image showing the target person and the imaging time at which each image showing the target person was taken.
1. An image monitoring system having a plurality of cameras and a server, wherein the server performs:
an abandoned item detection process which detects an abandoned item shown in an image taken by any of the plurality of cameras;
a person identification process which, in the case where the abandoned item has been detected, processes a sequence of images received from the camera that has taken the image from which the abandoned item has been detected in reverse order of time series from a point in time when the abandoned item was discovered so as to specify the image directly before the abandoned item had been placed, or processes the sequence of images in order of time series from a past point in time where it is clear that there is no abandoned item so as to specify the image directly after the abandoned item had been placed, and identifies at least one person shown in the specified image as a target person;
a person search process which searches for an image showing the target person from among the images respectively taken by the plurality of cameras;
a result display process which outputs on a screen a display showing the movement of the target person on the basis of the camera that took each image showing the target person and an imaging time at which each image showing the target person was taken; and
a personal feature value acquisition process which acquires a personal feature value including a face feature value and the worn clothing feature value of the target person on the basis of the specified image, wherein the person search process searches for the image showing the target person on the basis of the personal feature value, and
wherein, the person search process has a plurality of weighing patterns where a weight given to each element of a clothing feature value is differentiated, the person search process executing the trials of the search by applying the respective weighing patterns, selecting the weighing pattern that makes it easy to distinguish the resulting target person from other people, and executing the search by applying the selected weighing pattern,
wherein the result display process outputs on the screen a display showing a route which the target person has moved on a map or a display showing images of the target person that are aligned in order of time series for each camera as the display showing the movement of the target person, and displays a route which the target person passes thereafter, which is predicted on the basis of a past route which the target person had moved or a present route which the target person is moving.
2. The image monitoring system according to claim 1 , wherein the person identification process identifies a person who is at the closest distance from the abandoned item as the target person in the case where a plurality of people have been detected from the specified image.
3. The image monitoring system according to claim 1 , wherein the person identification process identifies a person who is at a distance of a predetermined threshold value or less away from the abandoned item or several people who are at a closer distance than others as the target people in the case where a plurality of people are detected from the specified image.
4. The image monitoring system according to claim 1 , wherein the person identification process identifies a person who is shown together with the abandoned item for the longest period of time is identified as the target person in the case where a plurality of people have been detected from the specified image.
5. The image monitoring system according to claim 1 , wherein in the case where the face of the target person is not shown in the specified image, the person identification process selects an image of the same person that clearly shows the face from among the images several frames before or after the specified image that have been taken by the same camera in order to acquire a facial feature value.
6. An image monitoring method performed by a server comprising the steps of:
detecting an abandoned item shown in an image taken by any of a plurality of cameras;
processing a sequence of images received from the camera that has taken the image from which said abandoned item has been detected in reverse order of time series from a point in time when the abandoned item was discovered in order to specify the image directly before the time when the abandoned item had been placed, or processing the sequence of images in order of time series from a past point in time where it is clear that there was no abandoned item in order to specify the image directly after the time when the abandoned item was placed, and identifying at least one person shown in the specified image as a target person;
acquiring a personal feature value including a face feature value and a worn clothing feature value of the target person on the basis of a specified image, wherein the person search process searches for the image showing the target person on the basis of the personal feature value,
searching for an image showing the target person from among the images respectively taken by the plurality of cameras; and
outputting on a screen a display showing the movement of the target person on the basis of the camera that took each image showing the target person and an imaging time at which each image showing the target person was taken, and
outputting on the screen a display showing a route which the target person has moved on a map or a display showing images of the target person that are aligned in order of time series for each camera as the display showing the movement of the target person, and displaying a route which the target person passes thereafter, which is predicted on the basis of a past route which the target person had moved or a present route which the target person is moving,
wherein, a plurality of weighing patterns where the weight given to each element of a clothing feature value is differentiated are prepared, and
the searching step executes the trials of the search by applying the respective weighing patterns, selects the weighing pattern that makes it easy to distinguish the resulting target person from other people, and executes the search by applying the selected weighing pattern.