PROBLEM TO BE SOLVED: To provide a finger shape estimation device for enabling finger shape estimation with high accuracy even though a person to be photographed moves freely and even though the person grabs something with a hand, and brings the palm into contact with other objects, such as the floor and wall, a desk and another person.SOLUTION: An RGB camera is attached to the back of a hand to image the back of the hand from a front slanting upward direction between the thumb and the index finger, and the bending and extension of the thumb, the index finger and the middle finger are accurately estimated. A plurality of pieces of image data of a hand or a moving object corresponding to the hand when the hand or the moving object is previously made to perform the same movement as a predetermined movement are associated with joint angle information on the hand collected previously in accordance with the plurality of pieces of image data for storage the hand or the moving object is made to perform desired movement movement image data of the hand or the moving object are acquired in time series during the movement image data corresponding to the movement image data are collated from the plurality of image data stored in an image correspondence joint angle storage device in time series and the joint angle information corresponding to the collated image data are outputted to perform finger shape estimation.COPYRIGHT: (C)2015,JPO&INPIT【課題】被撮影者が自由に動き回っても、手に何かを握ったり、掌を、床や壁、机、あるいは他者といった他物体に接触させても、高精度での手指形状推定ができるようにする手指形状推定装置を提供する。【解決手段】RGBカメラを、手の甲側に装着し、母指と示指のあいだの前方斜め上から手の甲を撮像し、母指、示指、中指の屈曲伸展を正確に推定する。事前に、手または手に対応する動作物に所定の動作と同じ動作をさせたときの手または動作物の複数の画像データと、複数の画像データに対応して事前に収集された手の関節角度情報とを対応付けて記憶し、手または動作物に所望の動作を行わせ、その際に手または動作物の動作画像データを時系列で取得し、動作画像データに対応する画像データを画像対応関節角度記憶装置に記憶している複数の画像データから時系列で照合し、照合した前記画像データに対応する前記関節角度情報を出力して手指形状推定を行う。【選択図】図1
