Moving object detection method and moving object detection apparatus

• US 8,363,902 B2
• Filed: 12/29/2010
• Issued: 01/29/2013
• Est. Priority Date: 07/29/2009
• Status: Expired due to Fees

• Current Assignee: Sovereign Peak Ventures, LLC (Dominion Harbor Enterprises, LLC)
• Original Assignee: Panasonic Corporation (Panasonic Holdings Corporation)
• Inventors: Nobori, Kunio, Iwasaki, Masahiro, Nakata, Mikiya
• Primary Examiner(s): Bella, Matthew
• Assistant Examiner(s): HEIDEMANN, JASON E
• Application Number:
  US12/980,903
• Publication Number:
  US 20110091074A1
• Time in Patent Office: 762 Days
• Field of Search: None
• US Class Current: 382/107
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• 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

Abstract

A moving object detection method includes: extracting NL long-term trajectories (NL≧2) over TL pictures (TL≧3) and NS short-term trajectories (NS>NL) over TS pictures (TL>TS≧2), using movement trajectories; calculating a geodetic distance between the NL long-term trajectories and a geodetic distance between the NS short-term trajectories (S205); calculating an approximate geodetic distance that is a geodetic distance between the NS movement trajectories over the TL pictures, based on the calculated geodetic distance between the long-term trajectories and geodetic distance between the short-term trajectories (S206); and performing segmentation based on the calculated approximate geodetic distance (S207).

Claims

1. A moving object detection method for detecting a moving object in video by performing segmentation on all or part of the moving object in the video, said moving object detection method comprising:

extracting NL long-term trajectories (NL≧

2) and NS short-term trajectories (NS>

NL) using movement trajectories each of which includes corresponding points over two or more pictures included in the video, the long-term trajectories being movement trajectories over TL pictures (TL≧

3), the short-term trajectories being movement trajectories over TS pictures (TL>

TS≧

2) included in the TL pictures;

calculating a geodetic distance between the NL long-term trajectories and a geodetic distance between the NS short-term trajectories;

calculating an approximate geodetic distance that is a geodetic distance between NS movement trajectories over the TL pictures, based on the geodetic distance between the long-term trajectories and the geodetic distance between the short-term trajectories calculated in said calculating of a geodetic distance; and

performing segmentation by segmenting a group of the movement trajectories as a single region, based on the approximate geodetic distance calculated in said calculating of an approximate geodetic distance.

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

wherein in said calculating of an approximate geodetic distance, a partial short-term trajectory is extracted based on the geodetic distance between the short-term trajectories, and the approximate geodetic distance is calculated based on the extracted partial short-term trajectory, the partial short-term trajectory being a group of the short-term trajectories.

3. The moving object detection method according to claim 1,

wherein in said calculating of an approximate geodetic distance, the approximate geodetic distance is calculated by performing linear interpolation on the geodetic distance between the long-term trajectories based on the geodetic distance between the short-term trajectories.

4. The moving object detection method according to claim 3,

wherein in said calculating of an approximate geodetic distance, the approximate geodetic distance is calculated by performing weighted linear interpolation so that a weight becomes larger as the geodetic distance between the short-term trajectories is smaller.

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

wherein in said calculating of a geodetic distance,

(i) a linear distance between the movement trajectories is calculated, (ii) a non-linearized distance is calculated by transforming, into an infinite value, a part of the linear distance between the movement trajectories, and (iii) a shortest distance among a plurality of distances of paths is calculated as the geodetic distance, the paths extending from one of two movement trajectories to the other of the two movement trajectories and passing through another movement trajectory as a relay point.

6. The moving object detection method according to claim 1,

further comprising:

capturing pictures included in the video; and

performing detection, on the pictures, of corresponding points between two pictures to calculate movement trajectories each including corresponding points over two or more pictures,

wherein in said extracting, the long-term trajectories and the short-term trajectories are extracted using the movement trajectories calculated in said performing detection.

7. The moving object detection method according to claim 6,

further comprising performing image processing on the video captured in said capturing pictures so that the video is displayed in a different mode for each of the regions segmented in said performing segmentation, and outputting the video after the image processing.

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

wherein in said performing image processing, NS approximate long-term trajectories are calculated, based on the approximate geodetic distance calculated in said calculating of an approximate geodetic distance and the NL long-term trajectories and the NS short-term trajectories extracted in said extracting, and video including the calculated approximate long-term trajectories is output, the NS approximate long-term trajectories being interpolated with a virtual long-term trajectory.

9. The moving object detection method according to claim 1,

further comprising estimating motion of the moving object by:

calculating a representative movement trajectory that represents the region, using movement trajectories included in the region segmented in said performing segmentation; and

estimating that the region moves according to the calculated representative movement trajectory.

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

wherein in said extracting, the long-term trajectories are extracted by changing the TL according to the number of extracted long-term trajectories.

11. The moving object detection method according to claim 10,

wherein in said extracting, the long-term trajectories are extracted by increasing the TL when the number of extracted long-term trajectories exceeds a predetermined first upper limit.

12. The moving object detection method according to claim 10,

wherein in said extracting, the long-term trajectories are extracted by decreasing the TL when the number of extracted long-term trajectories falls below a predetermined first lower limit.

13. The moving object detection method according to claim 1,

wherein in said extracting, the short-term trajectories are extracted by changing the TS according to the number of extracted short-term trajectories.

14. The moving object detection method according to claim 13,

wherein in said extracting, the short-term trajectories are extracted by increasing the TS when the number of extracted short-term trajectories exceeds a predetermined second upper limit.

15. The moving object detection method according to claim 13,

wherein in said extracting, the short-term trajectories are extracted by decreasing the TS when the number of extracted short-term trajectories falls below a predetermined second lower limit.

17. A program recorded on a non-transitory computer-readable recording medium, for detecting a moving object in video by segmenting all or part of the moving object in the video, which, when loaded into a computer, allows the computer to execute each step included in the moving object detection method according to claim 1.

16. A moving object detection apparatus which detects a moving object in video by performing segmentation on all or part of the moving object in the video, said moving object detection apparatus comprising:

a distance calculating unit configured to extract NL long-term trajectories (NL≧

2) and NS short-term trajectories (NS>

NL) using movement trajectories each of which includes corresponding points over two or more pictures included in the video, and to calculate a geodetic distance between the NL long-term trajectories and a geodetic distance between the NS short-term trajectories, the long-term trajectories being movement trajectories over TL pictures (TL≧

3), the short-term trajectories being movement trajectories over TS pictures (TL>

TS≧

2) included in the TL pictures;

a distance interpolation unit configured to calculate an approximate geodetic distance that is a geodetic distance between NS movement trajectories over TL pictures based on the geodetic distance between the long-term trajectories and the geodetic distance between short-term trajectories calculated by said distance calculating unit; and

a segmentation unit configured to perform segmentation by segmenting a group of the movement trajectories as a single region based on the approximate geodetic distance calculated by said distance interpolation unit.