IMAGE RECOGNITION APPARATUS, IMAGE RECOGNITION METHOD, AND GESTURE OPERATION APPARATUS

• Main Authors: TSUDA YUKA, OKUDA NORITAKA, NAKAMURA TAKEHIRO
• Format: Patent
• Language: English; Japanese
• Published: 21.05.2015
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

PROBLEM TO BE SOLVED: To improve accuracy for recognizing a gesture operation in response to disturbance light or a motion of an object.SOLUTION: An image recognition apparatus according to the present invention includes: an image selection unit that determines which image type a picked-up image photographed in an environment in which an irradiation device is periodically turned on and off belongs to, a light-on picked-up image when the irradiation device is turned on or a light-off picked-up image when the irradiation device is turned off; a motion-vector calculation unit that calculates a motion vector of an object using a plurality of picked-up images in the same image type; a differential-image generation unit that generates a corrected picked-up image by moving a position of the object included in the picked-up image on the basis of the motion vector, and that generates a differential image between the picked-up image and the corrected picked-up image; an object recognition unit that calculates features of a shape and a motion of the object from the differential image, and that outputs a recognition result indicating types of the shape and a hand motion of the object on the basis of the features; and an operation determination unit that determines an operation content from the object recognition result, and that generates and outputs a command indicating the determined operation content. 【課題】 外乱光や被写体の動きに対するジェスチャー操作の認識精度を向上させることを目的とする。【解決手段】 本発明に係る画像認識装置は、周期的に点灯及び消灯する環境下で撮影された撮像画像を、照射装置が点灯しているときの点灯時撮像画像及び消灯しているときの消灯時撮像画像のいずれの画像種類であるか判定する画像選択部と、画像種類が同じである複数の撮像画像を用いてオブジェクトの動きベクトルを算出する動きベクトル算出部と、撮像画像に含まれる被写体の位置を動きベクトルを基に移動した補正撮像画像を生成し、撮像画像と補正撮像画像の差分画像を生成する差分画像生成部と、差分画像から被写体の形状及び動きの特徴を算出し、特徴に基づいて被写体の形状及び手の動きの種類を示す認識結果を出力する被写体認識部と、被写体の認識結果から操作内容を決定し、決定した操作内容を示すコマンドを生成し出力する操作決定部とを設けた。【選択図】 図１