| 摘要 |
PROBLEM TO BE SOLVED: To realize identification and recognition matched to human sensitivity and to realize highly accurate recognition at an inexpensive calculation cost. SOLUTION: A gene is initialized (S1), an individual is evaluated (S2), and whether an evaluation value is within a specified value is determined (S3). When the evaluation value is outside the prescribed value, processing is repeated (S2, S3 and S4) in which the gene is crossed by an actual value GA, and whether the evaluation value is within the specified value is determined, and when the evaluation value is placed within the specified value, a queue A at this time is determined as a mapping function (S5). When the mapping function is determined, a mapping point of a learning image is calculated, its results are recorded in a database. At a recognition operation, it is possible to perform highly accurate recognition matched to human sensitivity at an inexpensive calculation cost by using the predetermined mapping function to map recognition images onto an identification space and outputting an image with the highest degree of approximation as a recognition image. COPYRIGHT: (C)2004,JPO
|
