Three-dimensional plant leaf mapping and segmentation using kinect camera

Automatic measurement of plant leaves is one of the key issues for vision guided robot operations in agricultural practices. In this paper, we present a study on plant leaf segmentation and 3D reconstruction of the leaf images using a time of flight camera. Depth and color images of plants are acquired in greenhouse by the low cost RGB-D camera and accordingly the depth and color data are utilized to extract plant leaves from the complicated background images. Initially, mean shift segmentation is conducted to obtain the connected components according to the depth information. Objects showing discontinuous distance are separated in this process, such as plants and the ground. Background images are subsequently removed by examining vegetation color of the segmented areas. Single leaves without occlusions are accurately segmented and 3D mapping is implemented after the background removed.