Three-Dimensional Reconstruction And Size Parameter Extraction Of The Plant Fruit And Leaf Based On Xtion

This paper is a study in the field of the 3d information acquisition, as extension section of national natural science fund project (60971115), detection system based on structure light sensor for three-dimensional reconstruction and the calculation of the plant leaf and fruit were discussed.Three-dimensional visual model of the plant leaf and fruit has important significance in analysising the growth law of the plant. By using the 3-D digital model of the plant leaf and fruit, the feature of type of plant varieties can be analysed, plant adaptation can be evaluated, cultivation management can be optimized, production can be analyzed, and dimension parameter can be measured.In addition, the three-dimensional digital model of the fruit and leaves of p1 ants or agricultural is an important prerequisite for agricultural robot in detectin g the environment. But so far the general studies just only aim at the plant leaf and fruit to use stereo vision system, three dimensional digitalization machine, scanning mode, optical precision instruments and rules for 3D reconstruction. However, in process of reconstruction based on stereo vision system, the stereo matching needs interaction to make sure correct features of the leaf and fruit. By using the method of scanning mode, optical precision instruments and three dimensional digitalization machine, the high-accuracy three-dimensional model can be acquired. But the above three methods cost is relatively high and it is difficult to be practice applied in agricultural production. And it is very difficult or impossible to use rule-based method to reconstruct morphological structure actually. So we adopt the xtion to realize the reconstruction of the fruit and leaf and calculate the size parameters. The main research contents and results are as follows.(1) The three-dimensional reconstruction of the fruitage. Multi-view point cloud images of the fruitage were acquired by using Xtion. Fruitage with different shapes and different textures, such as carambola, honeydew, eggplant, pineapple, maize and watermelon, were selected as the experimental samples to reconstruct their geometry models from cloud point data. The results showed that the three-dimensional reconstruction of the fruitage with different shapes and different textures basically reflected the natural morphology and the texture feature of the fruitage. The study can provide a new method for the reconstruction of the fruitage.(2) The acquisition of the 3-d information of the plant leaf in situ and the reconstruction of the plant leaf. The method of the acquisition of the 3-d information of the plant leaf in situ and the reconstruction of the plant leaf was developed. All cloud point data were acquired from a self-made leaf area acquisition system including a low-cost,3D camera Xtion sensor. The target leaf was recognized using distance judgment evidence and downsampled using a VoxelGrid filter. The cloud points filtered were reconstructed by means of Greedy Projection Triangulation. The results indicate that the target leaf could be segmented precisely using the method in this paper and the result of the three-dimensional reconstruction of the extracted leaf basically reflected the natural morphology of the target leaf.(3) The calculation of the morphology of the fruitage based on point clouds. The three-dimensional reconstruction of the fruitage with different shapes and different textures, such as carambola, honeydew, eggplant, pineapple and watermelon, were used to calculate the dimension parameter used solidwork and meshlab. The results indicate that the errors of volume, overall parameters and other morphological parameters were less than 10%. The study can provide a new method for the research on the calculation of the fruitage.(4) The leaf area in natural growth. Heron’s formula was used to calculate the leaf area based on the result of the reconstruction of the plant leaf. In order to verify the reliability of Leaf area calculation method, the proposed method was applied to measure different leaves representing varied shapes and sizes, such as Liriodendron, Emerald Duke and ping tree at a distance of less than 60 mm from the leaf. The measured error was within 3% compared with digital image measurement methods when the digital image processing method can be used as a reliable criterion. The method is applicable and reliable in determination of leaf area for modeling the growth of plants and plant physiology research.