| 摘要 |
PURPOSE:To enable secure and easy recognition of necessary information from picture patterns extracted by characters, graphics and a TV camera without using a format controlling system by providing a constitution recognizing means for performing image pattern constitution obtained after patterns to be recognized are quantized based on a prepared knowledge data base. CONSTITUTION:A constitution recognizing part 13 extracts characteristics such as complexity, size of black patterns, arrangement, repetitiveness, etc., from an image pattern, takes a general view and outputs pattern characteristics in the specific area and constitutional recognition results S consisting of said preceding patterns' addresses are imputted to an output buffer 14. A recognition controlling part 15 controls respective recognition parts 16a-16z by using said results S. Said 16a-16z output image patterns' recognition results to an edition controlling 17 by setting these data as code data Ra-Rz. The controlling part 17 confirms the contents of respective recognition results Ra-Rz according to said results S, and edits based on prepared knowledge data, by which necessary recognition results B are edited with respect to the image pattern and are outputted as the final response.
|
