To easily retrieve the image of a specific object from a plurality of images of an object whose image has been picked up by a camera installed on a passage road or the like, and to improve retrieval precision.
This image retrieval method is configured to fetch the image of an object such as a vehicle or a person whose image has been picked up by a camera (1) installed on a passage road or the like, and to generate a plurality of projection maps by projecting the image of the object by changing a camera visual point to a plurality of shape models, and to generate a front image by changing the camera visual point to the front face of the shape model based on each projection map, and to define the shape model corresponding to the front image whose bilaterial symmetry is the highest as the optimal shape model, and to select the corresponding visual point as the optimal camera visual point (105), and to generate an image for retrieval by projecting the projection map generated by projecting the image of the object from the optimal camera visual point to the optimal shape model to a preliminarily set camera visual point to convert the visual point (105), and to register the image for retrieval in the retrieval database, and to provide it for the retrieval of the object (107).
COPYRIGHT: (C)2008,JPO&INPIT
Tadaaki Kitamura
▲高▼橋 一哉
Shigetoshi Sakimura
Yutaka Sano
Tetsuya Yamazaki
JP2000039306A | ||||
JP2002056374A |