Machine Learning-Based System Automates Street Sign Detection and Classification

Geospatial scientists at RMIT University have used deep learning to produce an autonomous system for detecting traffic signs on Google Street View images. The system could allow local governments and traffic managers to monitor street signs and identify those in need of replacement or repair.

The scientists trained a custom object-detection model to detect and classify “Stop” and “Give Way” signs from images captured at intersection approaches. A photogrammetry approach was applied to calculate the approximate location of each detected sign in 2D geographical space. The experiments conducted on the road network area of study recorded a detection accuracy of 95.63% and a classification accuracy of 97.82%. The system was able to record the precise geolocation of signs from the 2D images. The researchers said that footage from other sources could be fed into the system, such images from cameras attached to municipal vehicles.

Object detection system for detecting and classifying road signs using GSV, RMIT University.

Object detection system for detecting and classifying road signs using GSV, RMIT University.

The system identifies and locates stop signs. Courtesy of RMIT University.

Newly located and classified street signs that are identified by the RMIT system could be combined with relevant spatial data to create an asset management system. By leveraging Google Street View, the RMIT system offers an economic way for towns and cities to build useful street sign computer vision data sets.

The researchers said that their system, which combines Google Street View with a geographical information system (GIS), is scalable to any level of street sign classification project. The system is based on free and open-source technology.

“Tracking these signs manually by people who may not be trained geoscientists introduces human error into the database,” researcher Andrew Campbell said. “Our system, once set up, can be used by any spatial analyst — you just tell the system which area you want to monitor and it looks after it for you.”

The research was published in Computers, Environment and Urban Systems (https://doi.org/10.1016/j.compenvurbsys.2019.101350).

QPC Lasers Inc. - QPC Lasers is LIDAR 4-24 MR

Published: June 2019

Glossary

artificial intelligence: The ability of a machine to perform certain complex functions normally associated with human intelligence, such as judgment, pattern recognition, understanding, learning, planning, and problem solving.
deep learning: Deep learning is a subset of machine learning that involves the use of artificial neural networks to model and solve complex problems. The term "deep" in deep learning refers to the use of deep neural networks, which are neural networks with multiple layers (deep architectures). These networks, often called deep neural networks or deep neural architectures, have the ability to automatically learn hierarchical representations of data. Key concepts and components of deep learning include: ...
machine learning: Machine learning (ML) is a subset of artificial intelligence (AI) that focuses on the development of algorithms and statistical models that enable computers to improve their performance on a specific task through experience or training. Instead of being explicitly programmed to perform a task, a machine learning system learns from data and examples. The primary goal of machine learning is to develop models that can generalize patterns from data and make predictions or decisions without being...

Browse Cameras & Imaging, Lasers, Optical Components, Test & Measurement, and more.