In the shape identification of the three dimensional object using a vision sensor, a system unaffected by the translation and the rotation of the object is required. The authors have developed a system for quickly identifying the shape of a 3-D object, using space encoding and a pattern transformation of image data. In this system, a dot-matrix-like pattern of bright spots is projected onto the object area, utilizing the optical diffraction effect produced by a fiber grating (FG) device. A spot projection pattern, which is obtained by the vision sensor, is transformed into a pattern of invariance to the translation of the object. The system is insensitive to the translation and the rotation of the object by training a multilayered neural network, using the transformed pattern.
This paper describes the method of shape identification of the 3-D object, using the FG vision sensor, and the configuration of the system. The results of the experiments, which were performed to verify the effectiveness of the proposed method. are discussed.
View full abstract