<i>Monitor geomatical information of plant by reconstruction 3D model based on Kinect V2</i>

Abstract. Traditional methods of geometrical shape analysis of plants most are based on 2D images. But this way is restrained by image position, sensors‘ prices, plants‘ density (e.g. overlapping problems) etc. With the introduced of RGB-D sensor, we can build plants‘ 3D model in little cost of time and money. Since the sensor, Microsoft Kinect V2, is inexpensive, which can be accessible and efficient for real-time plant monitoring in greenhouse. The 3D sensor equipped with a ToF (Time of Flight) light, is capable of capturing distance information, intensity and amplitude data in a single shot. By using the Kinect V2 as a sensor, this experiment shot RGB and RGB-D photo simultaneously in overhead position of plants, inside the house. Then we will reconstruct the 3D model and estimate the geometrical information, like biomass, leaf cover area, height, of the plants. Another destructive manual measurement done after the modeling to get the real geometrical information as compare. When analyzing accuracy of vegetables‘ parameters that gain from 3D model we can find that parameters, can be estimated at a receivable range. With the accuracy have been improved, automation operation in greenhouse can also be promoted.