| 摘要 |
PROBLEM TO BE SOLVED: To recognize a degrees of a risk to a variety of external environments by enabling a system to autonomously learn an experience in an actual environment. SOLUTION: Risk information is extracted by a risk information extraction part 4 from vehicle information (information on operations by drivers) inputted from a vehicle data input part 3, and teacher information (magnitudes and types of risks) is prepared. Characteristic information such as age information, movement information, and color information is extracted by an image characteristic amount extraction part 5 from image data from an image input part 2, and these pieces of information are held as an N-dimensional vector. A state recognizing part 6 converts the thus-obtained N-dimensional characteristic amount vector into a one-dimensional state. A risk recognizing part 7 obtains the correlation between a state obtained by the state recognizing part 6 and teacher data prepared by the risk information extraction part 4, and learns and recognizes a risk of the state. Then, the recognized risk is outputted from a risk information output part 8 visually or audibly. COPYRIGHT: (C)2009,JPO&INPIT
|
