Key Technologies Of Apple Tree Canopy Illumination Distribution And Growth Process Digitization Based On Point Cloud

Apple tree is an important economic plant.China is a big country of apple production and consumption in the world,but its excellent production rate and export rate are low.The morphological information and related physiological and biochemical parameters during the growth period of apple are indispensable important information to support the research on fruit tree cultivation and the digitalization and visualization technology has important theoretical significance and application value for scientific and precise field management during apple growth period.However,there are still some problems in fruit trees digitization field:high-precision point cloud scanning equipment is expensive,and tree point cloud data is difficult to be acquired unharmed and intact;The reconstructed tree model has low precision and botany characteristics;Most of growth process simulation research are based on rules with a low sense of reality and there are few studies on the interaction of environmental factors such as illumination.’Shanfu No.6’apple tree was chose as research object for the study of 3D reconstruction based on point cloud data,canopy light distribution prediction and growth simulation based on the technology chart of theoretical analysis,visualization simulation and experimental research,the main work is as follows:(1)Tree point cloud data acquisition and preprocessing scheme was completed.Depth data and color data were captured through 3D laser scanner Kinect V2.0 and software development kit which is supported by Microsoft company from double face.According to the distribution of noise point cloud,this paper used methods of passthrough filter and statistical filter to de-noising.Registration was divided into two steps:initial registration and fine registration.Artificial markers method was applied to initial registration which make initial position is reasonable.IRLS-ICP algorithm was applied to fine registration which make point cloud registration error is slight and registration error is 3.25cm.Data simplification through voxelization method which could maintain shape features.Data acquisition and preprocessing provide data foundation for tree model 3D reconstruction.(2)An angle-constrained space colonization algorithm was proposed based on skeleton information reconstruction strategy with preprocessed point cloud data.Based on the skeleton information,generalized cylinder is used to reconstruct the tree branch model.Finally,leaf model is added according to phyllotaxis,and a convenient reconstruction method of leaf model based on single image was proposed.The reconstruction results showed that the proposed method can simulate 3D morphological structure of tree realistically and showed the topological structure of trees well.Angle-constrained space colonization algorithm can obtain continuous skeleton information,and the reconstruction method has good universality and robustness.When the standard deviation of Gaussian noise is less than 0.02 or the simplification rate is less than 70%,it has little impact on the reconstruction results.Reconstructed tree models were plausible to the real-world trees and reconstruction error was less than 7.5%.(3)This article proposed a random forest prediction model based on canopy profile shadow feature and point cloud color feature.Using"slice method"to cut canopy model every0.1m on vertical direction,then using POV-Ray renderer to render shadows layer after layer,meanwhile,using light meter to obtain illumination intensity data every 0.1m from top to bottom consistently,and the random forest network that input data is color feature of every layer and output data is the relative illumination intensity was built as the apple tree canopy illumination distribution prediction model.The experiment results show that proposed method can predict the illumination distribution accurately.The average determination coefficient R~2between true value and predicted value is 0.853,and average MAPE is 27.1%.Random forest regression model can be used as a efficient method for prediction of canopy illumination distribution,and can provide reference for fruit tree pruning,plastic research.(4)Based on the theory of virtual organs,the apple tree growth cycle was divided into four parts:branch growth simulation,flowering simulation,leaf growth simulation and fruit expansion simulation.A point cloud boundary constraint-reversible semi-markov chain model was proposed based on the markov model,which simulate branches growth.The flowering period was simulated by the plane bidirectional deformation algorithm and leaves growth was simulated by distance-constrained allometric growth theory.Vernier caliper was used to continuously measure the middle and transverse diameters of apple fruits.Based on measured data,the growth curve of apple fruits was fitted,and fruit expansion process was simulated by 3D morphing technology.Meanwhile,the classic Faster Rcnn method was improved by non-maximum suppression algorithm based on three features.The original recognition rate is 84.1%and improved recognition rate is 91.3%.Based on the one-to-one mapping relationship between depth information and color information,the 3D coordinates of fruit were located.Finally,based on the cantilever beam mechanical model,the growth process is optimized.