MOVING OBJECT DETECTION APPARATUS AND MOVING OBJECT DETECTION METHOD

• US 20110255747A1
• Filed: 06/28/2011
• Published: 10/20/2011
• Est. Priority Date: 12/28/2009
• Status: Active Grant

• Current Assignee: Panasonic Corporation (Panasonic Holdings Corporation)
• Original Assignee: Panasonic Corporation (Panasonic Holdings Corporation)
• Inventors: Iwasaki, Masahiro, Nobori, Kunio, Komoto, Ayako
• Granted Patent: US 8,582,822 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06T 2207/30196 Human being; Person
• CPC Class Codes: G06T 2207/30241 Trajectory
• CPC Class Codes: G06T 7/215 Motion-based segmentation
• CPC Class Codes: G06T 7/223 using block-matching
• CPC Class Codes: G06V 40/25 Recognition of walking or running movements, e.g. gait recognition

Abstract

A moving object detection apparatus includes: an image input unit which receives a plurality of pictures included in video; a trajectory calculating unit which calculates a plurality of trajectories from the pictures; a subclass classification unit which classifies the trajectories into a plurality of subclasses; an inter-subclass approximate geodetic distance calculating unit which calculates, for each of the subclasses, an inter-subclass approximate geodetic distance representing similarity between the subclass and another subclass, using an inter-subclass distance that is a distance including a minimum value of a linear distance between each of trajectories belonging to the subclass and one of trajectories belonging to the other subclass; and a segmentation unit which performs segmentation by determining, based on the calculated inter-subclass approximate geodetic distance, a set of subclasses including similar trajectories as one class.

Claims

1. A moving object detection apparatus which detects a moving object in video by performing segmentation on an image to specify all or part of regions of the moving object in the video, said moving object detection apparatus comprising:

an image input unit configured to receive input of a plurality of pictures included in the video;

a trajectory calculating unit configured to calculate a plurality of trajectories by detecting, for each block including at least one pixel that is included in one of the received pictures, a motion of the block between two pictures included in the video, and concatenating detected motions for the plurality of pictures;

a subclass classification unit configured to classify the calculated trajectories into a plurality of subclasses each of which is a set of similar trajectories;

an inter-subclass approximate geodetic distance calculating unit configured to calculate, for each of the subclasses, an inter-subclass approximate geodetic distance which represents similarity between the subclass and another one of the subclasses, using an inter-subclass distance that is a distance including a minimum value of a linear distance between each of trajectories belonging to the subclass and one of trajectories belonging to the other one of the subclasses; and

a segmentation unit configured to perform segmentation by specifying, based on the calculated inter-subclass approximate geodetic distance, a set of subclasses including similar trajectories as a class.

2. The moving object detection apparatus according to claim 1,

wherein said inter-subclass approximate geodetic distance calculating unit is configured to calculate, for each of the subclasses, as the inter-subclass approximate geodetic distance between the subclass and the other one of the subclasses, a distance that is a sum of:

an intra-subclass distance representing similarity between trajectories belonging to the subclass;

an intra-subclass distance representing similarity between trajectories belonging to the other one of the subclasses; and

the inter-subclass distance between the subclass and the other one of the subclasses.

3. The moving object detection apparatus according to claim 2,

wherein said inter-subclass approximate geodetic distance calculating unit is configured to calculate, for each of the subclasses, as the intra-subclass distance of the subclass, an average, mode, or median of a linear distance between the trajectories belonging to the subclass.

4. The moving object detection apparatus according to claim 2,

wherein said segmentation unit is further configured to calculate, as the inter-subclass approximate geodetic distance, a distance calculated by normalizing, using the intra-subclass distances, the inter-subclass approximate geodetic distance between two subclasses.

5. The moving object detection apparatus according to claim 1,

wherein said inter-subclass approximate geodetic distance calculating unit is configured to calculate, when the minimum value is larger than a predetermined threshold, the inter-subclass approximate geodetic distance by performing nonlinearization for transforming the inter-subclass approximate geodetic distance into an infinite value.

6. The moving object detection apparatus according to claim 5,

wherein said inter-subclass approximate geodetic distance calculating unit is configured to compare an inter-subclass approximate geodetic distance between a first subclass and a second subclass, and a sum of an inter-subclass approximate geodetic distance between the first subclass and a third subclass and an inter-subclass approximate geodetic distance between the third subclass and the second subclass, and to calculate a smaller value as the inter-subclass approximate geodetic distance between the first subclass and the second subclass.

7. The moving object detection apparatus according to claim 1,

wherein said segmentation unit is configured to perform segmentation by specifying, based on temporal variation in the inter-subclass approximate geodetic distance, the set of subclasses including similar trajectories as the class representing a region.

8. The moving object detection apparatus according to claim 7,

wherein said segmentation unit is configured to perform segmentation by determining that the two subclasses belong to a same class, when a value indicating the temporal variation in the inter-subclass approximate geodetic distance between the two subclasses is equal to or less than a predetermined threshold for the temporal variation.

9. The moving object detection apparatus according to claim 7,

wherein said segmentation unit is configured to perform segmentation by determining that the two subclasses belong to different classes, when a value indicating the temporal variation in the inter-subclass approximate geodetic distance between the two subclasses is larger than a predetermined threshold for the temporal variation.

10. The moving object detection apparatus according to claim 1,

wherein said inter-subclass approximate geodetic distance calculating unit is configured to determine, for each of the subclasses, as the inter-subclass distance between the subclass and the other one of the subclasses, an average of a predetermined number of distances selected, in ascending order, from among distances between the each of the trajectories belonging to the subclass and the one of the trajectories belonging to the other one of the subclasses.

11. The moving object detection apparatus according to claim 1,

wherein said inter-subclass approximate geodetic distance calculating unit is configured to determine, for each of the subclasses, as the inter-subclass distance between the subclass and the other one of the subclasses, an average of distances equal to or less than a predetermined threshold, from among distances between the each of the trajectories belonging to the subclass and the one of the trajectories belonging to the other one of the subclasses.

12. The moving object detection apparatus according to claim 1,

wherein said subclass classification unit is configured to classify the trajectories into the subclasses by classifying similar trajectories into a same subclass, from among the calculated trajectories.

13. The moving object detection apparatus according to claim 12,

wherein said subclass classification unit is configured to classify the trajectories into the subclasses by repeatedly performing, predetermined times or until the trajectories are classified into a predetermined number of subclasses, processing for classifying, into a same subclass, trajectories having highest similarity from among the calculated trajectories.

14. The moving object detection apparatus according to claim 1,

wherein said subclass classification unit is configured to classify the trajectories into the subclasses by segmenting, into a plurality of regions, at least one of the pictures received by said image input unit, using at least one of edge information, color information, and brightness information of the at least one of the pictures, and classifying, into a same subclass, trajectories passing through each of the regions resulting from the segmentation.

15. The moving object detection apparatus according to claim 1, further comprising

an output unit configured to perform image processing on at least one of the pictures received by said image input unit such that each of the regions specified by the segmentation performed by said segmentation unit is displayed in a mode different from a mode in which another one of the specified regions is displayed, and to output a picture including the region on which the image processing has been performed.

16. The moving object detection apparatus according to claim 1, further comprising

a recording and transmission unit configured to write onto a recording medium or transmit to an outside via a transmission path, at least one of the pictures that is assigned with a result of the segmentation performed by said segmentation unit, the pictures being received by said image input unit.

17. The moving object detection apparatus according to claim 1, further comprising

a prediction unit configured to calculate a representative trajectory that represents the class specified by said segmentation unit from among the trajectories included in the class, and to predict, according to the calculated representative trajectory, a motion of the moving object by predicting that a region corresponding to the class moves between pictures.

18. A method of detecting a moving object in video by performing segmentation on an image to specify all or part of regions of the moving object in the video, said method comprising:

receiving input of a plurality of pictures included in the video;

calculating a plurality of trajectories by detecting, for each block including at least one pixel that is included in one of the received pictures, a motion of the block between two pictures included in the video, and concatenating detected motions for the pictures;

classifying the calculated trajectories into a plurality of subclasses each of which is a set of similar trajectories;

calculating, for each of the subclasses, an inter-subclass approximate geodetic distance which represents similarity between the subclass and another one of the subclasses, using an inter-subclass distance that is a distance including a minimum value of a linear distance between each of trajectories belonging to the subclass and one of trajectories belonging to the other one of the subclasses; and

performing segmentation by specifying, based on the calculated inter-subclass approximate geodetic distance, a set of subclasses including similar trajectories as a class.

19. A program for detecting a moving object in video by performing segmentation for specifying all or part of regions included in the moving object in the video, said program causing a computer to execute steps included in the method of detecting a moving object according to claim 18.

20. A non-transitory computer-readable recording medium,

wherein data having a structure including header information and body information is recorded on said non-transitory computer-readable recording medium, and

the header information includes:

a size of pictures included in video;

the number of the pictures included in the video;

the number of trajectories obtained from the video; and

the number of subclasses each of which is a set of similar trajectories, and

the body information includes;

as many pictures as indicated by the number of the pictures included in the video;

data representing as many trajectories as indicated by the number of the trajectories obtained by the video;

data representing the subclass to which each of the trajectories belongs; and

an inter-subclass approximate geodetic distance representing, for each of the subclasses, similarity between the subclass and another one of the subclasses.

21. A moving object detection data generating apparatus comprising:

an image input unit configured to receive input of a plurality of pictures included in the video;

a trajectory calculating unit configured to calculate a plurality of trajectories by detecting, for each block including at least one pixel that is included in one of the received pictures, a motion of the block between two pictures included in the video, and concatenating detected motions for the plurality of pictures;

a subclass classification unit configured to classify the calculated trajectories into a plurality of subclasses each of which is a set of similar trajectories;

an inter-subclass approximate geodetic distance calculating unit configured to calculate, for each of the subclasses, an inter-subclass approximate geodetic distance which represents similarity between the subclass and another one of the subclasses, using an inter-subclass distance that is a distance including a minimum value of a linear distance between each of trajectories belonging to the subclass and one of trajectories belonging to the other one of the subclasses; and

a data output unit configured to output data including at least the inter-subclass approximate geodetic distance calculated by said inter-subclass approximate geodetic distance calculating unit.

22. A moving object detection apparatus which detects a moving object in video by performing segmentation on an image to specify all or part of regions of the moving object in the video, said moving object detection apparatus comprising:

a data input unit configured to receive an input of data including at least an inter-subclass approximate geodetic distance representing similarity between each of trajectories belonging to a subclass which is a set of similar trajectories in the video and one of trajectories belonging to another one of the subclasses that is different from the subclass and is a set of similar trajectories in the video; and

a segmentation unit configured to perform segmentation by specifying, based on the inter-subclass approximate geodetic distance received by said data input unit, a set of subclasses including similar trajectories as one class.