遺伝的アルゴリズムによる2次元画像からの3次元物体姿勢認識

概要

論文の詳細を見る
This research is concernedd with a new technology of the image recognition of 3D object using genetic algorithms. Recent1y， autonomous robots such as mobile robots are widely studied. However，as the image obtained from the vision system is 2D，the recognition technology of 3D object is necesary for the robots to grip the object. In this research，a new technique is proposed to recognize both the position and the posture of 3D object from 2D image by applying GA. First，search model is converted to the data of 2D image. That is，each parameters of the rotation angles of x， y， z axis，the reduction rate，and the amount of the movement are decided by using GA expressed by the gray code. Then，2D data is obtained by using the afain conversion from these paramenters. Second，the recognition expriment is performed between search model obtained by GA and rhe original image，and the recognition result is obtained. Fina1ly，through these experiments，it can be shown that the new technology proposed here is sufficient1y valid for the position and the posture recognition of 3D objects from 2D image.
福井大学工学部の論文