An Improved Stereo Matching Algorithm Applied to 3D Visualization of Plant Leaf

Digital agriculture has become a hot research areas in recent years. However, low cost and high accuracy are always paradox in agriculture. In this paper, a digitization technology is introduced and utilized to reconstruct virtual plant leaf model by using computer graphics and machine vision technologies. In order to obtain the accurate 3D point cloud of the object with low cost, a binocular stereo vision system is created to realize digitization of the scene. In order to adapt to the complex environmental, an improved stereo matching algorithm based on gradient and census is proposed, which is robust to uneven illumination condition. In the fourth section, the digitization technology is used to acquire the 3D point cloud of the plant leaf, which can depict the structure accurately. Then a 3D Delaunay triangulation algorithm and texture-mapping technology are applied to reconstruct a 3D life-like leaf based on the point data with real leaf image. The digitization and visitation technology can been used to create the library of plant leaves, which provide reliable samples for building the growth model of plant.