CarGo

## Theme name Smart Mobility Challenge by Hitachi Vantara

Team name

CarGo

Project details

We develop a next generation tunnel auto toll system in order to encourage usage and speed up tunnel auto payment process. Thus, to make tunnel traffic more efficient and effective.

Background

Auto toll system is a proven and effective technology to reduce traffic congestion. In Hong Kong, the auto toll system (ETC) is accepting 43% of the vehicles crossing the tunnel everyday (according to transport department monthly report 2018 July), despite having much fewer booth compare to manual toll collection booth. However, when we go deeper into the reports, back in 2001, the ETC is already accepting 39% of the total vehicles. There seem to be a reluctance for drivers to apply the auto toll system. The current auto toll system require drive to pay a monthly admin fee and install a RFID tag onto their vehicle. We think that this is the key reason.

Inspiration

Recent years, AI and e-wallet have been developed rapidly. In 2013, Google make use of the neural network system to to recognise house numbers from street views and achieving human-level accuracy. We think that combining this two technology, we can create a cheaper and simpler auto toll system.

What it does

Provide simplified process for driver to use tunnel auto toll. Our system consists of two parts: the tunnel side detection system and user mobile application.

The first part used to detect which car is entering the tunnel, we have to setup a camera in toll plaza to capture car license plate. Then charge the driver automatically.

The second part is mobile application on drivers' phone. It receive payment notice and manage tunnel payment history. Our system supports linkage to wide range of e-wallets and payment systems for auto-charging tunnel fee when using. It is also very easy to register the service, drivers just need to provide the proof of their identity and car license.

How I built it

Use the latest opensource tool, TensorFlow, OpenCV, Node.js, Vue.js (Progressive Web App)

collect useful dataset from google, university open data (e.g. SUN dataset from MIT)
train our neural network with large amount of dataset set, so it can recognise license plate and background
build our backend and fron-end to cooperate with the recognition system