Human detection and tracking apparatus, human detection and tracking method, and human detection and tracking program
A human detection and tracking apparatus prevents errors in a size of a person and a location of body parts between actual image data and a tracking result. Human frame detecting section detects, from first image data, a human frame as a region having high possibility of presence of a human, based on human feature data representing a feature of an entire human body. Body part frame location determining section determines a body part frame in the first image data, based on part feature data illustrating a feature of a body part of the human and a part frame determined as a region having high possibility of presence of a body part of the human in second image data previous to the first image data. Body part frame location correcting section corrects, based on the human frame, a location of the part frame determined in the first image data.
1. A human detection and tracking apparatus comprising:
a human frame detecting section that detects, from first image data in a current frame, a human frame as a region having a high possibility of presence of a human, based on human feature data representing a feature of an entire human body;
a body part frame location determining section that determines a plurality of body part frames in the first image data, based on (i) body part feature data illustrating a feature of a body part of the human and (ii) a plurality of body part frames determined as regions having a high possibility of presence of each body part of the human in second image data in a preceding frame; and
a body part frame location correcting section that corrects, based on the human frame, each location of the plurality of body part frames determined in the first image data so as to match the human frame,
wherein the body part frame location determining section (i) sets a calculation range based on the plurality of body part frames determined in the second image data and the human frame detected in the first image data, (ii) calculates, in the calculated range, body part likelihood based on the body part feature data for each region having a size identical to a size of the body part frames determined in the second image data, and (iii) determines a region selected based on the body part likelihood as a body part frame in the first image data.
2. The human detection and tracking apparatus according to claim 1 , further comprising
a simplified posture estimation section that estimates a posture of the human in the first image data, based on the location of the plurality of body part frames determined by the body part frame location determining section, wherein, when the posture of the human estimated by the simplified posture estimation section matches the human frame, the body part frame location correcting section corrects each location of the plurality of body part frames based on the human frame, the location being determined in the first image data.
3. The human detection and tracking apparatus according to claim 2 , wherein, when the posture of the human estimated by the simplified posture estimation section does not match the human frame, the body part frame location correcting section does not correct each location of the body part frames determined in the first image data.
4. The human detection and tracking apparatus according to claim 1 , further comprising
a body part frame setting section that sets a default body part frame in the first image data based on the human frame, when the body part frame is not determined in the second image data.
5. The human detection and tracking apparatus according to claim 4 , wherein
the body part frame setting section divides the human frame into regions in a vertical direction, and sets each of the divided regions as the default body part frame.
6. A human detection and tracking method comprising:
detecting, from first image data in a current frame, a human frame as a region having a high possibility of presence of a human, based on human feature data representing a feature of an entire human body;
determining a plurality of body part frames in the first image data, based on (i) body part feature data illustrating a feature of a body part of the human and (ii) a plurality of body part frames determined as regions having a high possibility of presence of each body part of the human in second image data in a preceding frame; and
correcting, based on the human frame, each location of the plurality of body part frames determined in the first image data so as to match the human frame,
wherein the determining step (i) sets a calculation range based on the plurality of body part frames determined in the second image data and the human frame detected in the first image data, (ii) calculates, in the calculated range, body part likelihood based on the body part feature data for each region having a size identical to a size of the body part frames determined in the second image data, and (iii) determines a region selected based on the body part likelihood as a body part frame in the first image data.
7. A human detection and tracking program stored on a non-transitory computer-readable medium causing a computer to execute processing, the processing comprising:
detecting, from first image data in a current frame, a human frame as a region having a high possibility of presence of a human, based on human feature data representing a feature of an entire human body;
determining a plurality of body part frames in the first image data, based on (i) body part feature data illustrating a feature of a body part of the human and (ii) a plurality of body part frames determined as regions having a high possibility of presence of each body part of the human in second image data in a preceding frame; and
correcting, based on the human frame, each location of the plurality of body part frames determined in the first image data so as to match the human frame,
wherein the determining step (i) sets a calculation range based on the plurality of body part frames determined in the second image data and the human frame detected in the first image data, (ii) calculates, in the calculated range, body part likelihood based on the body part feature data for each region having a size identical to a size of the body part frames determined in the second image data, and (iii) determines a region selected based on the body part likelihood as a body part frame in the first image data.
8. The human detection and tracking method according to claim 6 , further comprising
estimating a posture of the human in the first image data, based on the location of the plurality of body part frames determined by the determining step, wherein, when the posture of the human estimated by the estimating step matches the human frame, the correcting step corrects each location of the plurality of body part frames based on the human frame, the location being determined in the first image data.
9. The human detection and tracking method according to claim 8 , wherein, when the posture of the human estimated by the estimating step does not match the human frame, the correcting step does not correct each location of the body part frames determined in the first image data.
10. The human detection and tracking method according to claim 6 , further comprising
setting a default body part frame in the first image data based on the human frame, when the body part frame is not determined in the second image data.
11. The human detection and tracking method according to claim 10 , wherein
the setting step divides the human frame into regions in a vertical direction, and sets each of the divided regions as the default body part frame.
12. The human detection and tracking program according to claim 7 , further comprising
estimating a posture of the human in the first image data, based on the location of the plurality of body part frames determined by the determining step, wherein, when the posture of the human estimated by the estimating step matches the human frame, the correcting step corrects each location of the plurality of body part frames based on the human frame, the location being determined in the first image data.
13. The human detection and tracking program according to claim 12 , wherein, when the posture of the human estimated by the estimating step does not match the human frame, the correcting step does not correct each location of the body part frames determined in the first image data.
14. The human detection and tracking program according to claim 7 , further comprising
setting a default body part frame in the first image data based on the human frame, when the body part frame is not determined in the second image data.
15. The human detection and tracking program according to claim 14 , wherein
the setting step divides the human frame into regions in a vertical direction, and sets each of the divided regions as the default body part frame.