| 摘要 |
PROBLEM TO BE SOLVED: To perform robust object detection for an appearance change due to illumination environments or an attitude change and for an environment change under complex background. SOLUTION: A change object detector 7 generated by a learning part 6 using change regions and no-change regions in a video is used as a front stage to achieve candidate region extraction of a detection object being robust to the appearance change and the environment change. As a rear stage, a pre-learning object detector 10 is assembled. Finally, a connection likelihood determining part 13 determines a threshold and detects the object using a connection likelihood obtained by combining the object likelihood calculated by an object likelihood calculating part 9 using the change object detector 7 and the object likelihood calculated by an object likelihood calculating part 11 using the object detector 10. COPYRIGHT: (C)2011,JPO&INPIT
|
