Motivation
Inner city districts often get swamped in car traffic from the suburbs even if their residents do not use cars themselves a lot. Manual counting of vehicle categories, e.g., car, bicycle, and pedestrians, which is still applied by the city of Vienna is time-consuming and expensive but can already be recognized well using networks developed for autonomous driving. To measure the extent of drive-thru traffic in specified areas, car license plates can be scanned when entering and exiting these areas, using pole-mounted cameras. The recognized plate codes can be hashed immediately, permitting to determine what percentage of registered motor vehicles pass that area within a specific amount of time, while still preserving drivers' privacy.
Description
This requires testing and applying vehicle detection and license plate recognition algorithms, possibly adapting them to pole-mounted views (from overhead view) plus designing, constructing, and mounting an energy-autonomous device that can film traffic during daylight hours outdoors on traffic poles.
Tasks (depending on PR/BA/DA and number of students)
- Design the device hardware, e.g., Raspberry PI, AI module, USB output solar module with power bank, suitable camera, and robust housing
- Test an open-source software pipeline for edge hardware, e.g., yolo vehicle classification, open-alpr license plate recognition, hashing + counting that can recover from errors as much as possible without manual intervention
- Evaluation of the quality of classification, plate recognition, and statistics of drive-thru percentage depending on the amount of time allowed
Requirements
- Knowledge of English (source code comments and final report have to be in English)
- Experience in Linux, Raspberry platform, python, and usage of neural networks is a plus
Environment
A bonus of €500/€1000 if completed to satisfaction within an agreed time-frame of 6/12 months (PR/BA or DA)