摘要
HOG算法在行人检测和手势识别中有很好的试验效果,是一种优秀的特征描述符。通过对HOG算法做适当变形处理,使其应用于Kinect的RGB和深度图像,以此描述目标物体的3D特征信息,通过libSVM分类器训练模板,并进行物体的分类和识别。试验结果表明:利用RGB和深度信息的HOG识别系统比只利用二维信息的识别系统有着更高的识别准确率。
HOG algorithm has a very good experimental performance in pedestrian detection and gesture recognition, and is a good feature descriptor. It may be used in the RGB and depth image of Kinect sensor by making an appropriate deformation to the HOG algorithm. Then it is used to describe the 3D features of the objects. The objects＂ classification and recognition may be achieved by training the object model through the libSVM classifier. The experiments show that it will have higher recognition accuracy in HOG recognition system by using RGB and depth information than using two dimensional information only.
出处
《农业科技与装备》
2013年第1期35-37,共3页
Agricultural Science & Technology and Equipment
