Title:
DATA PROCESSING UNIT
Document Type and Number:
Japanese Patent JPS60126772
Kind Code:
A
Abstract:
PURPOSE:To classify excellently a class by selecting a pivot at simple clustering to a maximum value placed nearly at the center of the class from the beginning, and changing a threshold level as a parameter so as to decrease the number of times of revision. CONSTITUTION:X and Y coordinates relating points X1...XN extracted from a two-dimensional histogram and the level are stored in a memory 1-1 of a data processing unit. The memory 1-1 is referenced by a sequencing operating section 2, the level of the points X1...XN is compared, the levels are rearranged in order beginning from the largest in level order and the result is stored in a memory 1-2. Moreover, a class classifying operating section 3 obtains the Euclid distance of points Y1...YN sequenced by the operating section 2, the distance and the threshold value are compared and when the distance is the threshold value or below, the threshold value is revised. Moreover, the dispersion in each class classified by an evaluation criterion operating section 4 is evaluated to a minimum value in average, a threshold value revising section 5 revises the threshold value to a minimum value, the result of classification by the operating section 3 is obtained and stored in a memory 1-3 so as to decrease the number of times of revision.
Inventors:
MARUYAMA TSUGUHITO
Application Number:
JP23589883A
Publication Date:
July 06, 1985
Filing Date:
December 14, 1983
Export Citation:
Assignee:
FUJITSU LTD
International Classes:
G06T7/00; G06T5/40; G06T7/60; (IPC1-7): G06F15/62; G06K9/00; G06K9/36
Attorney, Agent or Firm:
Yutaka Morita