Shadow type identification for gait recognition using shadows

Abstract

\Using features acquired from the shadow cast by a walking person can be an alternative for gait recognition whenever the person’s body is occluded, such as when capturing images from an overhead position. However, the shadow, depending on the light source characteristics, can be cast as a blob with no distinguishing characteristics. Most state-of-the-art methods fail in the presence of such “diffused” shadows. Thus, this paper presents a novel method to identify the type of shadow cast by a person. The proposed method generates a histogram of the intensity ratio between foreground and background areas, whose analysis allows identifying the type of shadow cast by the person. The proposed method is very promising, achieving a 90% correct shadow type identification with the dataset tested.