Grass Detection Based on Color Features

Terrain analysis plays an more and more important role in the research of cross-country environment understanding, We collected lots of pictures which contain grass and classified the grass as green grass and yellow grass. With the research of grass distribution in different color spaces, we choose YC_bC_r color space which has the best aggregation for further research. We build a semi-circle model and an ellipse model to describe the distribution of green grass and build a parallelogram model and an ellipse model to describe the distribution of yellow grass. With the help of the models, we develop programs to detect the grass. Lastly, the proposed ellipse-based grass detection algorithm has shown an encouraging performance in grass detection experiments on a self-captured cross-country environment video.