Street sign recognition driver assist system

RaS:MAN Overview

Utilizing real time image recognition, street signage is identified for use in driving assistance.

Street sign information is analyzed against the vehicle's CAN/OBD-II to provide the driver with law adherence guidance.

A voice interface allows the user to query recent sign information. Multiple languages are supported, making it ideal support for drivers visiting from overseas.

Target Market
Personal automobiles, instrument panels, dashboard cameras, vehicles intended for rental car distribution
Target Device

RaS:MAN Features

RaS:MAN is a system to provide driving support using real time traffic sign recognition and a voice driven user interface. Traffic sign data is collected in real time with image recognition running on an embedded device. Driving status is then extrapolated by analyzing the current vehicle information and sign data, which the voice interface uses to provide the driver with useful warnings and information.

RaS:MAN stores the traffic sign history and provides driver support through it's voice interface. Sign data obtained through image recognition is also able to be combined with map data to provide a higher level of driving support.

RaS:MAN Functionality

Traffic sign recognition
High speed recognition with TensorFlow SSD (Single Shot MultiBox Detector)
Achieves 60 FPS using GPU/TPU acceleration hardware
Vehicle data collection
Collects vehicle information with CAN / OBD-II
Able to recognize speed, gear, steering, etc.
Voice recognition and speech
Voice interface and announcements in Japanese, English, Mandarin, Cantonese, and Korean
Interfaces using Microsoft Azure or AWS
Lightweight footprint
Runs on a Raspberry Pi4