In this paper, we present an innovative approach to passenger flow monitoring for light rail transportation networks. We propose a distributed system based on two main concepts. On each vehicle in the transportation network, a set of sensors is used to count people at a given place. On a cloud-based server, a data synchronization and storage system aggregates the data sent from all vehicles and provides a global view of the transportation network. The contribution, with respect to the state of the art, of our approach is twofold. First, the proposed distributed architecture is able to reduce the system global cost via its flexibility and ease of deployment, since the main part of the system is onboard each vehicle and not fixed at stations or track sections. Second, the novel vision-based passenger counting approach guarantees high levels of reliability in the estimation of the number of people in a given area, and the ability to provide real-time data on the global transportation network. Experimental results demonstrate the validity and the advantages of the proposed approach, paving way to future uses of the system as the base of additional network optimization modules for the global light rail transportation.