| 摘要 |
PURPOSE:To accurately extract a feature point of a work by calculating a calculation range by using point-row data indicated by a matrix and a characteristic matrix statistically processed. CONSTITUTION:When a cross sectional shape of a work 10 is measured by means of a cross sectional shape measuring sensor 11, a matrix-conversion operating section 15 converts the obtained point-row data of the cross section into matrix formation data. In a learning stage, a characteristic matrix-forming section 18 converts a calculation range inputted by each point-row data into matrix formation data to form a characteristic matrix. A statistics operating section 20 averages the characteristic matrices formed up to that time. The characteristic matrix is learned step by step in the experience as an intermediate item combining in the probability the point-row data with the calculation range. In an execution stage, a sample range calculating section 22 calculates a calculation sample rang based on the point-row data represented by the matrix and averaged characteristic matrix and a virtual point-calculating section 23 applies a prescribed calculation operation to the inputted calculation sample range, thereby obtaining the virtual point (feature point). |
