基于Kinect深度图像的人体识别分析

介绍了深度图像在模式识别中的研究现状及其在人体识别中的应用。针对目前普通相机拍摄的图像识别在光照、姿态、遮挡等因素影响下性能下降的问题,以微软推出的Kinect设备为平台,通过分析Kinect相机获取的深度图的特征,提出以综合点特征和梯度特征的局域梯度特征的方式来对人体部位区分判定,并以手肘为例作了简要论证。

Analysis of human identification based on kinect depth image

Nowadays somatosensory human-computer interaction devices have become hotspot applications in the field of digital media. These devices capture the depth images of players through several inner cameras and sensors, from which human skeletons can be extracted, and players' movement can be tracked and captured. We introduce the research status of depth image in the field of pattern recognition and application of human recognition. Since the recognition of images captured by common cameras shows poor performance in the influence of illumination, posture and overlap, based on the device of Microsoft Kinect, the features of depth images captured by Kinect cameras are analyzed. Then, local gradient features integrating point features and gradient features have been put forward to identify human body. Brief demonstration and analysis are given by taking elbow as an example.