OBJECT POSITION INFERENCE DEVICE, OBJECT POSITION INFERENCE METHOD, AND PROGRAM

摘要

Incorrect position inference due to a virtual image object is reduced in determining the position of an object by view volume intersection. An object position inference device comprises an object candidate acquisition means for inferring the position of an object by view volume intersection and acquiring an object candidate, an object candidate information storage means for storing information about the overlap of the object candidates in respect of the view volumes of the object obtained from a captured image, a co-movement degree acquisition means for acquiring an index which is the degree of co-movement representing the co-movement of the object candidates calculated from the number of view volumes of object candidates including an object candidate and another object candidate moving together when the object candidate moves, and an object candidate determining means for determining each object candidate as the object or a non-object according to the posterior probability on the basis of the acquired index and outputting the position of the object candidate determined as the object.