SimpleLinefindingForIGVC

Ken Cramer
May 3, 2012

Attach:kcramer_linefinding.doc

Simple Linefinding for IGVC

Overview

This project is an implementation of a simple technique to isolate and segment out the white features, namely lines, of an image. This is done via a series of simple image filters applied to an image using the openCV framework.

Photograph

screenshot of the algorithm run on a sample video

Concepts Demonstrated

  • Spatial image manipulation is used to isolate white features
  • Image transformation is used to obtain a bird's eye view of the area in front of the robot
  • Raycasting is used to isolate interesting features closest to the robot

Innovation

Past approaches to this problem by the UML robotics team have relied upon complicated filtering techniques ending with a hough transform to characterize the lines present. This project uses a series of heuristics to classify the image, and then a customized mixing of color channels based upon that classification. After filtering, perspective transform and raycasting the image is then directly converted into an occupancy grid.

Technology Used Block Diagram