| KEYWORD |
Transport Research for Innovation and Sustainability (TRIS)
Automated Passenger Counting (APC) through video image recognition
keywords AUTOMATED PASSENGER COUNTING, VIDEO CAMERA, VIDEO PROCESSING
Reference persons CRISTINA PRONELLO
Research Groups Transport Research for Innovation and Sustainability (TRIS)
Thesis type EXPERIMENTAL - DESIGN
Description Passenger counting inside public transport systems uses various methods. There are several devices in the market, mainly using optical systems as stroboscopic cameras or video cameras. The analysis of the market shows that these systems are declared to have a high accuracy in laboratory test, but this is not the case in real field. Analysing pros and cons and the accuracy on the field - getting the data from the transport companies - will allow to understand how, starting from a videocamera existing on the market, the counting can be improved using machine learning and computer vision.
Use a video camera available on the market (provided by the teacher) and trying to implement a video image recognition algorithm, starting from what available in literature and, through test on site (e.g., in GTT vehicles), understand the routes of improvement. One of the outcomes will be also to use a not expensive videocamera and search the tradeoff between price of the system and accuracy. The final outcome will be a simple prototype embedding an algorithm that gives an accuracy good enough and that can work well together other systems, not based on an optical approach. The need is to find a trade-off between cost and accuracy, in an environment that can host more than one device for counting passengers.
Required skills Programming knowledge especially in Python, some knowledge of C will be beneficial but not mandatory. Some knowledge of Image recognition algorithms.
Deadline 05/11/2024
PROPONI LA TUA CANDIDATURA