| Summary: | Cycling is increasingly popular as a sustainable urban mobility mode. Data can play a key role in the success of cycling promotion initiatives, by helping to understand cycling demand and assess the real impact of investments. However, creating a global view of the cycling activity of a city can be a major challenge, as there are no obvious sources from which to obtain the necessary cycling data. While there are now many bike counters in the market, their hardware and deployment costs severely limit the number of sensors that can be deployed and consequently the spatial coverage of the city. In this work, we explore the viability of a large-scale bicycle counting infrastructure for city-wide cycling analytics, which explores the trade-off between costs and spatial/temporal coverage. The proposed solution uses a set of temporary low-cost video-based counters, which can flexibility be rotated among multiple counting locations. To understand the viability of the approach we developed a prototype counter, where we tested two video processing techniques: OpenCV and Yolo. Results suggest that the overall approach could indeed support a low-cost and universal bike counting functionality, as long as delayed access to data is acceptable. Even though this is not envisioned as a general bike counting solution, it may provide a smart way to approach the complex issue of universal city-wide bike counting.
|
|---|