RoadBuster

Problem

Low accuracy of the existing license plate identification system has resulted in significant manual effort on the part of Traffic Police officers. How might we improve the accuracy of identifying license plate details, so that Traffic Police officers can do other higher value-added work?

Solution

RoadBuster uses Optical Character Recognition (OCR) to identify licence plate numbers and reduce the time taken for Traffic Police officers to confirm red light and speeding offences, making the roads safer for everyone. The prototype is a back-end admin interface and dashboard for officers to confirm licence plate details from traffic camera images.

See our slides here:

See the full deck on Google Slides.

User Research

We have done a preliminary check with Traffic Police, and have gotten their feedback that the prototype, if it works as intended, will allow them to save time and manpower. We are pending a wider trial with Traffic Police. 

Roadbuster was also shared at Demo Day during Hackathon 2024, and another agency has expressed interest in collaborating with us.

Further Collaborations

Further potential developments on Roadbuster include:

1. Refining the existing image processing workflow with Traffic Police

2. Improving the model with more images 

3. API integration with existing systems

Other than Roadbuster, Traffic Police is keen to work with us on two further areas:

1. Shortening the current waiting time to send out notices, which currently takes weeks

2. Streamlining the appeals process 

We’re in talks with Traffic Police to develop these ideas further. 

Team

Roadbuster was brought to you by the ERP X team, a multi-disciplinary team of engineers, designers, product ops and policy officers.

Left to right: Daryl Chan, Justyn Oh, Stephanie Siow, Blake Gong, Christabel Png, Mike Chen, Samuel Koh (not in picture)

Posters