To provide a means which not only improves robustness by excluding an influence of hiding but also improves recognition precision by optimizing the position and posture of a three-dimensional object, in a three-dimensional object recognition device which recognizes a three-dimensional object by features such as a contour in a two-dimensional image.
The three-dimensional object recognition device includes: a camera which photographs a three-dimensional object; a means which projects the three-dimensional object on a camera image while changing the position and posture of the three-dimensional object and calculates coordinates of respective projection points and directions of edges; a means wherein a lookup table generated on the basis of calculated them is stored; a means which creates pyramid images; a means which extracts edges with respect to a pyramid image having the lowest resolution; a means which creates a directional distance map; a means which maps respective projection points on the directional distance map; a means which evaluates the position and posture; and a means which optimizes the position and posture.
COPYRIGHT: (C)2010,JPO&INPIT
JP2002185695 | IMAGE READER |
JP2003008832 | METHOD FOR POSITIONING READ SENSOR, AND IMAGE READER |
JP2014182433 | ROAD STATE COMPREHENSION APPARATUS |
JP200993611A |