Human activity recognition-based path planning for autonomous vehicles
- PDF / 978,913 Bytes
- 8 Pages / 595.276 x 790.866 pts Page_size
- 34 Downloads / 237 Views
ORIGINAL PAPER
Human activity recognition-based path planning for autonomous vehicles Martin Tammvee1 · Gholamreza Anbarjafari2,3,4 Received: 15 August 2020 / Revised: 15 August 2020 / Accepted: 5 October 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract Human activity recognition (HAR) is a wide research topic in a field of computer science. Improving HAR can lead to massive breakthrough in humanoid robotics, robots used in medicine and in the field of autonomous vehicles. The system that is able to recognise human and its activity without any errors and anomalies would lead to safer and more empathetic autonomous systems. During this research work, multiple neural networks models, with different complexity, are being investigated. Each model is re-trained on the proposed unique data set, gathered on automated guided vehicle (AGV) with the latest and the modest sensors used commonly on autonomous vehicles. The best model is picked out based on the final accuracy for action recognition. Best models pipeline is fused with YOLOv3, to enhance the human detection. In addition to pipeline improvement, multiple action direction estimation methods are proposed. Keywords Neural networks · Self-driving car · Object detection · Human detection · Human action detection · Path planning
1 Introduction Human activity recognition (HAR) is a wide field of study dedicated on identifying the specific movement or action of a person based on acquired data. Data can be gathered by multiple different sensors, depending on field of usage for HAR [1]. Most common activities that are tracked are walking, standing and sitting. Actions can be more specific if model needs to be used in more narrow field, for example, medicine. This work has been partially supported by the Estonian Centre of Excellence in IT (EXCITE) funded by the European Regional Development Fund. The authors also gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan X Pascal GPU.
B
Gholamreza Anbarjafari [email protected] Martin Tammvee [email protected]
1
Cleveron Ltd., Reinu tee 48, 71020 Viljandi, Viljandi County, Estonia
2
iCV Research Lab, Institute of Technology, University of Tartu, 50411 Tartu, Estonia
3
PwC Advisory, Helsinki, Finland
4
Department of Electrical and Electronic Engineering, Hasan Kalyoncu University, Gaziantep, Turkey
This research work contributes on finding the best model of HAR for self-driving cars and improving them with stateof-an-art techniques. Second part of the research work is validating a new data set gathered by the most modern sensors in the field of self-driving car. Third part of the research work is adding new features to the researched models as well as proposing various methods estimating humans movement direction in videos. Autonomous cars can be allowed into public areas only when they are completely safe to humans. As the resources on the self-driving cars are limited, the procedure cannot be computationally expensive, while at the same time it has to run
Data Loading...
