Heliaus (ECSEL)

Heliaus (ECSEL)

With the emergence of new forms of personal mobility, the automotive industry is facing a drastic change. Photonic technology – one of the six Key Enabling Technology (KET) recognized by the European Commission – plays a crucial role in meeting automotive needs and expectations. With all sensors combined, limitations are still expected and will limit availability of assistance or autonomous vehicle.

Taking benefit of Long Wave Infra-Red bandwidth (LWIR) in which all object radiates energy depending only on its structure and temperature, the HELIAUS project aims to deliver breakthrough perception systems. HELIAUS will develop smart thermal perception systems that will detect Long Wave Infra-Red light both for in-cabin passengers monitoring and for the car surrounding. HELIAUS will improve object classification of the automotive sensor suite in all light conditions, provide redundancy and thus extend vehicle autonomy towards level 3 and beyond, operating 24/7.

Heliaus – Advanced Thermal Imaging Algorithms

Primary Researchers:
Muhammed Ali Farooq, Waseem Shariff

Supervisors:
Peter Corcoran

Description:
As part of the HELIAUS project, which aims to deliver breakthrough thermal perception systems that will detect Long Wave Infra-Red light both for in-cabin passengers monitoring and for the car surrounding, the core focus of this project is to improve object detection/classification of the automotive sensor suite in all light and weather conditions. Thus far, this work has involved developing various in-cabin driver monitoring applications including smart thermal gender classification system, synthetic data generation, face detection, eye tracking and facial landmarks detection system using advanced machine learning algorithms. Moreover, a set of object detection systems in thermal spectrum have recently been developed and trained for road-side environment monitoring systems using the state-of-the-art CNN-YOLO framework. Finally, work is on-going to deploy these advanced end-to-end neural architectures on resource-constrained and low-power edge devices for real-time onboard applications. 

Funded by:  This project has received funding from the ECSEL Joint Undertaking (JU) under grant agreement No 826131. The JU receives support from the European Union’s Horizon 2020 research and innovation programme and France, Germany, Ireland, Italy. Completed by local funding from the French region Auvergne Rhône Alpes.