INOVSAFE - In and out of vehicle threat identification and monitorization for traffic safety

The project is meant to advance existing scientific approaches in 2 different directions: extracting relevant road traffic features and data from a single camera sensor and monitoring ego-vehicle state from a single camera sensor coupled with inertial measurement data. The main objective of this project is to design and implement algorithms and solutions for in and out of vehicle threat identification and monitoring to increase traffic safety. The road and ego-vehicle perception will be implemented using computer vision and artificial intelligence. The system can offer relevant data to the driver, in order to prevent hazardous situations that lead to accidents. The project has the objective to implement algorithms that can accurately observe, measure, detect and track the road traffic scene using a single camera. Another objective is to monitor the ego-vehicle state using the same input data and also inertial measurement data in order to detect hazardous scenarios.

The expected original contributions are:
- developing artificial intelligence and computer vision algorithms to process sensorial data acquired using various cameras and inertial sensors
- designing and implementing an artificial neural network that is able to generate and extract multiple relevant features regarding the traffic scene from monocular images in a single step
- developing a computer vision algorithm and artificial neural network solution for monitoring the ego-vehicle state and hazardous road traffic scenarios

From the Technical University of Cluj-Napoca the following key persons:

• Dr.Eng. Razvan Itu - project manager
• Professor Radu Danescu - project mentor

The workplan of the project is organized in three phases, corresponding to the calendar years covered.

Phase 1 (April 2022 – Dec 2022):

• Initial prototype of the road traffic perception system. Traffic sensory data acquisition system and ego-vehicle monitoring. Dissemination of results and website for the project.

Phase 2 (Jan 2023 – Dec 2023):

• Perception system using vision and artificial intelligence capable of providing relevant information about the road traffic scene using monocular images and analyzing the ego of the vehicle and the driver. Dissemination of results.

Phase 3 (Jan 2024 – Mar 2024):

• Integration of perception systems in a single demonstrator application. Dissemination of results.

Conferences:

Razvan Itu, Radu Danescu, “On-Board Estimation of Vehicle Speed and The Need of Braking Using Convolutional Neural Networks”, In Proceedings of the 20th International Conference on Informatics in Control, Automation and Robotics - Volume 1 (ICINCO), 2023, SciTePress, pp. 600-607.

Razvan Itu, Radu Danescu, “Predicting Emergency Braking in Vehicles Using a CNN with Sequential Image and Velocity Data”, 2023 IEEE 19th International Conference on Intelligent Computer Communication and Processing (ICCP 2023), 2023, pp. 41-47.

ISI Journals:

Razvan Itu, Radu Danescu, “Fully Convolutional Neural Network for Vehicle Speed and Emergency Brake Prediction”, Intelligent Vehicle Sensing and Monitoring, Sensors, Vol. 24, No. 1, 2024, Art. No. 212.

Razvan Itu, Radu Danescu, “Part-Based Obstacle Detection Using a Multiple Output Neural Network”, Sensors, vol. 22, no. 12, p. 4312, Jun. 2022, doi: 10.3390/s22124312

Phase 1 report here.

Phase 2 report here.

Phase 3 report here.

FINAL report here.

Razvan Itu

Address:
Technical University of Cluj-Napoca
Computer Science Department,
Str. Memorandumului, Nr. 28, 400 114, Cluj-Napoca
Romania

Office: Baritiu str. 26, room 37
Phone: +40 264 401457

E-mail:
Razvan.Itu@cs.utcluj.ro

Full project name:
INOVSAFE - In and out of vehicle threat identification and monitorization for traffic safety

Contract:
PN-III-P1-1.1-PD-2021-0247

Funding:
Ministry of Research and Innovation, CNCS – UEFISCDI