摘要 |
PROBLEM TO BE SOLVED: To perform highly accurate recognition, using only with a luminance image, when recognizing an object by comparing a feature amount of the object constructed on a three-dimensional virtual space with a feature amount of an image generated by imaging the object in actual space. SOLUTION: A scene setting process section 11 sets a three-dimensional model scene stored in a database 2. An image generating process section 12 generates a distance image and a luminance image for each set scene. An image feature calculating process section 13 calculates the feature amount of the distance image and the feature amount of the luminance image. A correlation coefficient calculating process section 14 of the distance/luminance image feature amount calculates a correlation coefficient between the feature amount of the distance image and the feature amount of the luminance image. A luminance-to-distance image feature conversion expression constructing process section 15 creates a conversion table for converting the feature amount of the luminance image to the feature amount of the distance image using high correlation. COPYRIGHT: (C)2011,JPO&INPIT
|