| 摘要 |
PURPOSE:To enable a picture recognition with the aid of using a means which is compact in respect of both a hardware and a software by discriminating the data of a sampled point by using the formula of Maharanobis' generalized distance. CONSTITUTION:An object 101, the data of which is to be collected, is image-picked up, and a picture information is analyzed by a photoelectric converting part 102 into the color element information of a red, a green and a blue (RGB), and into the differentiation information of the RGB by a unit of a picture element, and collected. The color element information, a luminance information and the differentiation information are normalized by a variable normalizing part 103, and an aimed normalization information is stored in a memory part 105. Next, a mean vector and a variance-covariance matrix are calculated by a operation processing part 104 on the basis of these normalization informations, and the formula of Maharanobis' generalized distance for discriminating and analyzing is formed and stored in the memory part 105. At the time of the picture recognition, the object 101' to be recognized is image-picked up, and the obtained normalization information is classified and recognized by the operation processing part 104 by using the formula of Maharanobis' generalized distance, stored in the memory part 105, and a recognized result is displayed on an input/output device 106.
|
