发明名称 Object detection apparatus, object detection method, monitoring camera system and storage medium
摘要 An object detection apparatus comprising: an extraction unit adapted to extract a feature amount for each of partial regions that constitute image data; a storage unit adapted to store, as a background feature amount for each of the partial regions, a feature amount extracted in advance from background image data and does not include the target object; a first determination unit adapted to determine whether or not the partial region is an object region including the target object by comparing a current feature amount extracted from the partial region with the background feature amount; a deciding unit adapted to decide a parameter for each of the partial regions based on a plurality of results of determination; and a second determination unit adapted to determine whether or not the partial region is the object region based on the parameter and the results of determination.
申请公布号 US9098748(B2) 申请公布日期 2015.08.04
申请号 US201113267455 申请日期 2011.10.06
申请人 CANON KABUSHIKI KAISHA 发明人 Yano Kotaro
分类号 H04N7/18;G06K9/00 主分类号 H04N7/18
代理机构 Fitzpatrick, Cella, Harper & Scinto 代理人 Fitzpatrick, Cella, Harper & Scinto
主权项 1. An object detection apparatus that detects a target object from image data, the apparatus comprising: an image obtaining unit adapted to obtain image data by sequentially capturing an image in a predetermined direction from a predetermined viewpoint; a feature extraction unit adapted to extract a feature amount for each of partial regions that constitute the image data obtained by the image obtaining unit; a storage unit adapted to store, as a background feature amount for each of the partial regions, a feature amount extracted in advance by the feature extraction unit from background image data that has been obtained by the image obtaining unit and does not include the target object; a first object region determination unit adapted to determine, for each of the partial regions, whether or not the partial region is an object region including the target object by comparing a current feature amount extracted from the partial region by the feature extraction unit with the background feature amount; a statistic calculation unit adapted to calculate, for each of the partial regions, a statistic of determination within a period by the first object region determination unit, based on determination results by the first object region determination unit within the period; a parameter deciding unit adapted to decide a parameter for each of the partial regions based on a statistic of incorrect determination calculated by the statistic calculation unit within a parameter setting period in which the partial region does not include the target object but is incorrectly determined as the object region by the first object region determination unit; and a second object region determination unit adapted to determine, for each of the partial regions determined as the object region by the first object region determination unit within a detection period in which the object would be detected, whether the partial region is the object region or a disturbance, based on the parameter decided by the parameter deciding unit and the statistic of determination within the detection period calculated by the statistic calculation unit.
地址 Tokyo JP
您可能感兴趣的专利