TY - GEN
T1 - Hidden Human Target Detection Model Inspired by Physiological Signals
AU - Zhang, Lejing
AU - Wang, Yunlu
AU - Hu, Menghan
AU - Li, Qingli
N1 - Publisher Copyright:
© 2022, Springer Nature Singapore Pte Ltd.
PY - 2022
Y1 - 2022
N2 - The current object detection algorithms will give unsatisfactory performance on the task of detecting hidden human targets. Therefore, in the current work, we propose a physiological signals powered hidden human targets detection model. The new proposals generation algorithm considering the spatio-temporal interdependent physiological features is first proposed to generate suitable candidate boxes. To eliminate the oddly shaped candidate boxes, we introduce a priori knowledge by combining the candidate box size and aspect ratio. The skin detection model is used to further reduce the number of candidate boxes. The custom-made dataset is established to validate the performance of the proposed model. The model we built yields the accuracy of 64% and 44% for indoor and outdoor environments, respectively. Compared to the YOLOv4, in terms of indoor and outdoor scenes, the developed model shows 30% and 16% improvement in accuracy. The results of the ablation experiments show the effectiveness of each component of the model.
AB - The current object detection algorithms will give unsatisfactory performance on the task of detecting hidden human targets. Therefore, in the current work, we propose a physiological signals powered hidden human targets detection model. The new proposals generation algorithm considering the spatio-temporal interdependent physiological features is first proposed to generate suitable candidate boxes. To eliminate the oddly shaped candidate boxes, we introduce a priori knowledge by combining the candidate box size and aspect ratio. The skin detection model is used to further reduce the number of candidate boxes. The custom-made dataset is established to validate the performance of the proposed model. The model we built yields the accuracy of 64% and 44% for indoor and outdoor environments, respectively. Compared to the YOLOv4, in terms of indoor and outdoor scenes, the developed model shows 30% and 16% improvement in accuracy. The results of the ablation experiments show the effectiveness of each component of the model.
KW - IPPG
KW - Occluded human detection
KW - Physiological signal measurement
KW - Proposals generation
UR - https://www.scopus.com/pages/publications/85128917527
U2 - 10.1007/978-981-19-2266-4_18
DO - 10.1007/978-981-19-2266-4_18
M3 - 会议稿件
AN - SCOPUS:85128917527
SN - 9789811922657
T3 - Communications in Computer and Information Science
SP - 229
EP - 238
BT - Digital TV and Wireless Multimedia Communications - 18th International Forum, IFTC 2021, Revised Selected Papers
A2 - Zhai, Guangtao
A2 - Zhou, Jun
A2 - Yang, Hua
A2 - An, Ping
A2 - Yang, Xiaokang
PB - Springer Science and Business Media Deutschland GmbH
T2 - 18th International Forum of Digital Multimedia Communication, IFTC 2021
Y2 - 3 December 2021 through 4 December 2021
ER -