Study On Point Cloud Data Modeling Forthe Populus Standing Timber

The difficulty of the three-dimensional reconstruction of the standing timber point clouddata is to simplify complex and scattered mass measurement data, and the key technology is tocut a few corners of related algorithms. On the one hand, in regard to individual standing timbersample, the geometric shape of the populus trunk is extremely irregular and unstructured,branches and leaves grow optional and in dispersion, the position of the nodule is abrupt, andthe trunk measured data obtained by the laser scanner is extraordinary numerous and jumbled;and the other hand, even the same age within the same species, the difference of timberbiological traits between dissimilar individuals is huge. Information got by mining a smallnumber of individual sample point cloud data is difficult to promote directly to apply to theoverall samples of the forest for the error cannot be estimated. Expanding the sample sizeaccording to the number of the trees for the estimation of the error means multiplied increase ofthe mass measurement capacity. Therefore, the algorithm efficiency is crucial. The mainpurpose of this thesis is to improve the algorithm efficiency of processing the timber pointcloud data under the premise of guaranteed accuracy.In this thesis, after the pretreatment of the populus standing timber point cloud data,adopting, the trunk of the standing timber is fitted with the help of Bicubic Bezier SurfaceInterpolation Fitting Method. Although this method can get relatively good fitting data, it alsocause certain serious lacunae of the trunk data. On this basis, the thesis proposes the PipeReconstruct Transform Method which combines the Reconstruction Pipe Surface Method andRandom Hough Transform Detection Method. The fitting result of the method proposed obtainsthe whole timber trunk data, and it has superiority on the surface fitting of the timber data overBicubic Bezier Surface Interpolation Fitting Method. Then the thesis applies topologyinformation obtained by K-Nearest Neighbors Searching which has been simplified into featureextraction, in which some of algorithms related have been altered which are ‘step by step’ onapproaching reasonable feature points. The main research contents in this thesis are as follows:(1)In application and research on point cloud data from scholars at home and abroad recentyears, relatively, most geometric shapes of the research object are generally more regular andstructured, which does not apply to the standing timber point cloud data, for the geometry shapeof timber is extremely unstructured. This thesis preprocesses the entire point cloud from theperspective of the color information(R, G, B) with the characteristics of the standing timber—populus point cloud data. The artificial interference crude extraction algorithm based on RGBcolor information performances simple, targeted and operability. Experiments show that thismethod is a feasible pretreatment method of the standing timber modeling;(2)For the thesis of the populus main trunk, this thesis combines Reconstruction PipeSurface Method and Random Hough Transform Detection Method and propose Pipe Reconstruct Transform Method to implementing on the fitting of the main trunk, then contrastthe proposed method with the Bicubic Bezier Surface Interpolation Fitting Method. The trunkfitting precision of the Bicubic Bezier Surface Interpolation Fitting Method in this thesis is lessthan77.584%, while the Pipe Reconstruct Transform Method almost completes fitting theentire trunk. The proposed Pipe Reconstruct Transform Method gains the advantage over theBicubic Bezier Surface Interpolation Fitting Method in the trunk fitting problem;(3)Then the thesis simplifies the topology information acquisition algorithm of the pointcloud data, and then improved the feature extraction algorithm of local point cloud based on theobtained local topology information above. In the process of extracting main branches featuresof the populus point cloud data, putting forward Stepwise Approaching Method to gain thecollection of feature points. This method utmost ensure the local area characteristics of thepopulus point cloud model, and comparatively preserve the outline of the entire populus pointcloud model as well. The experimental results turn out that the simplification of the K-NearestNeighbors Searching and the established Stepwise Approaching Feature Extraction Method arepracticable.