This repository contains basic applications for traffic sign recognition and lane detection. The code is written in C# .NET and uses Emgu CV (3.1.0.2504) as wrapper for OpenCV.
A SURF (Speeded up robust features) based detection and recognition tool. It uses the homography with RANSAC to increase stability. A set of known signs is matched with candidates in the image.
A Haar based sign detector. Multiple Haar cascade files are included to detect a 90km/h speed limit sign (Belgian).
A lane detector based on white or yellow road markings. This code is conceptualy based on the Udacity 'Self-Driving Car Engineer Nanodegree' program.
- Download and install 'emgucv-windesktop_x64-cuda 3.1.0.x' from https://sourceforge.net/projects/emgucv/files/emgucv/3.1.0/
- Clone or download this repository
- Open TrafficComputerVision/Code/TrafficComputerVision/TrafficComputerVision.sln
- Update the references in the projects to your Emgu installation.
- Build for x64
All necessary test data can be found under TrafficComputerVision/Code/TestData/