立体形状の高速認識方式

Abstract

This paper comprehensively presents a method to recognize 3-dimensional objects within one second. Slit light data from a range finder and gray scale image from a TV-camera are used jointly. First, by subtracting the background image from the background with slit light image, an image containing only, the slit light is obtained. ±0.3 pixel accuracy in fixing the slit line position in the image is obtained by using a statistical method. The extracting time is 200 ms. Then syntax analysis for labeling objects and extraction of contour lines from the gray scale image line is reported. The threshold for processing the gray scale image is decided automatically by making use of the slit light. At last, the method to reconstruct 3-dimensional data from 2-dimensional data is presented. The coefficients between picture coordinates (I, J) and a spacial coordinates (x, y, z) are calculated using a least squares method. Totally, 400 ms for all execution time and ±0.2 mm in accuracy, are obtained.