A Feature Fusion Model For Person Identification Using Top-view Image

In this paper, Top-view images taken by the bird'seye view digicam can reduce the limitation of the camera installation position and solve the problem of the face in image being blocked. we count on people's clothes, hair color, and physique do now not exchange in a quick length of time. We propose a person identification method from a bird's-eye view image by fusing features. To obtain a high discrimination rate even with a small number of learning images, we extract effective features (1) HOG (2) gray level co-occurrence matrix (3) color histogram in the traditional computer vision field. And (4) VGG16 in deep learning field. And then we fused these features and applied them to train the SVM classifier. The effectiveness of the proposed approach used to be tested from many comparative experiments.