Plant Recognition by Integrating Color and Range Data Obtained Through Stereo Vision

We present a new method for recognizing plants automatically and nondestructively by measuring plants three-dimensionally, including height, width, and leaf area, which are important clues for determining a plant condition. We use only two cameras and small plant preparation. A pair of color images for a plant is obtained by a binocular stereo camera, then leaf and stalk areas are extracted from color images, range data is calculated from stereo images, three-dimensional reconstruction for the plant is generated by using the Delaunay triangulation and the plant is measured in the generated reconstruction. Two experimental results are shown for actual plants.