Multi-person Identification and Localization Algorithm Using RGB-D Image Segmentation
Xu Han, Jun Yan · 2022
In recent years, with the development of smart phones and intelligent monitoring systems, visual object detection and localization methods have been widely used. Since RGB-D based person identification and localization technology have achieved better performances, an algorithm based on deep learning and image segmentation is proposed in this paper. First, the raw images are preprocessed. Then, image segmentations are performed according to the results of person detecting by YOLOv3 network. For multi-person identification, the Haar cascade classifier is used to determine regions of face on the segmented color images and the Eigen Face method is used for classification. For person localization, the convolutional neural network is preformed regression learning based on the depth images of a single person. Experiments results show that the proposed algorithm performs better than other existing approaches.