Research On Reconstruction Of Plant Leaf By Depth Camera Kinect

With the rapid development of computer hardware technology and image processing algorithm, virtual reality technology is more and more widely used in people’s daily life, including science fiction films, somatic games, smart wear and so on. As an important part of the computer graphics fields, computer virtual plant modeling has been continuously exploring and innovating. From the classical plant modeling methods of L system and fractal system to the plant modeling method based on image, and then to the present plant modeling method based on three dimensional point cloud, the rebuilt models have been gradually improved in terms of running speed and sense of reality.In this paper, the leaves are made as the research object. A new algorithm named multiple density leaf reconstruction with limited details is provided on the basis of eyes recognition. Firstly, the leaf information through the depth camera of Kinect is received,which including the depth data and color data. Some relevant research methods have been put forward for solving a series of problems of scanning. Finally, the model of the leaf is got. The contents of relevant research are as follows:(1) The leaf point cloud model is obtained by using the Kinect. After the initial operation of dealing with noise and points match, a simplified method of point cloud reduction based on RGB is proposed. The method combines with traditional bounding box algorithm, which makes the point cloud color more natural and effectively separates the bonded lobed leaf.(2) Combining with implicit function surface reconstruction, the leaf is reconstructed by the Poisson algorithm based on the simplified point cloud of RGB. Comparing to the surface reconstruction of the original point cloud, the two models are analyzed for the similarity degree, the details of the reconstruction effect, the time and space of reconstruction performance.(3) A new algorithm named multiple density leaf reconstruction with limited details is provided on the basis of eyes recognition. The algorithm is different from traditional grid reconstruction. It generates the fuzzy surface by refining the density of points in order toreach the purpose of surface. Finally, the experiment shows that the result and speed of the new algorithm is better than mesh reconstruction in a certain extent.(4) By using the computer graphics theory of OpenGL library and Qt GUI programming, the system is grouped by the function of leaf point cloud simplification,surface reconstruction, multiple density with limited details reconstruction. Still more,according to the need of users, the system can be set the relevant parameters to show the results of experiment more intuitive and rich.