Real-time line crossing detection
SONAR|HES-SO
- Rithner, Aurélien
- Carrino, Francesco (Degree supervisor)
- Sion : Haute Ecole d'Ingénierie, HES-SO Valais-Wallis, 2023
Bachelor of Science (BSc) HES-SO ingénieur·e en Systèmes industriels: Haute Ecole d'Ingénierie, HES-SO Valais-Wallis, 2023
English
This report documents a project carried out as part of the "Summer School I" program within the Industrial Systems curriculum at HES-SO Sion. Over a three-week period, students are tasked with building remote-controlled vehicles for a race competition.
The objective is to develop a system that can automatically detect when a vehicle crosses track marking in real-time using cameras.
First, an analysis of the problem's specifications, the different hardware choices available and the different techniques used in this computer vision field was made.
Then, using Python and OpenCV, a prototype was developed with two major parts. The back end assuring image acquisition and processing as well as data extraction. The front end is a Django based web server that allows to store information in a database and display them on a web page.
Evaluations shows that the prototype achieved timing and crossing detection using traditional methods, while the front-end web page allows users to monitor the race information and watch the best times saved on a scoreboard.
A final proof-of-concept was designed and will be put to test at this year’s summer school race.
Therefore, a mock-up was built to test the system under similar conditions to those of the actual race.
Additionally, a machine learning based approach was explored as well. It required to generate virtual images of the race for the data set used to train the CNN model using Unity and Blender. With encouraging initial results, this approach was put on the side for time reasons.
The objective is to develop a system that can automatically detect when a vehicle crosses track marking in real-time using cameras.
First, an analysis of the problem's specifications, the different hardware choices available and the different techniques used in this computer vision field was made.
Then, using Python and OpenCV, a prototype was developed with two major parts. The back end assuring image acquisition and processing as well as data extraction. The front end is a Django based web server that allows to store information in a database and display them on a web page.
Evaluations shows that the prototype achieved timing and crossing detection using traditional methods, while the front-end web page allows users to monitor the race information and watch the best times saved on a scoreboard.
A final proof-of-concept was designed and will be put to test at this year’s summer school race.
Therefore, a mock-up was built to test the system under similar conditions to those of the actual race.
Additionally, a machine learning based approach was explored as well. It required to generate virtual images of the race for the data set used to train the CNN model using Unity and Blender. With encouraging initial results, this approach was put on the side for time reasons.
- Language
-
- English
- Classification
- Electronics
- Notes
-
- Haute Ecole d'Ingénierie Valais
- Systèmes industriels - Systemtechnik
- Infotronics
- hesso:heivs
- Persistent URL
- https://sonar.rero.ch/global/documents/328568
Statistics
Document views: 23
File downloads:
- TB_Aurélien Rithner rapportFinal.pdf: 29