3D Reconstruction Of Tree Crown Based On Unmanned Aerial Vehicle Aerial Image

As an important part of the aircraft, unmanned aerial vehicle (UAV) aerial hasbecome an important method to obtain space data. In the field of military and civilian,UAV becomes an irreplaceable means of surveying and mapping. The UAV aerial2Dimages without3D solid geometry data of the object, lack of intuitive, so in order toobtain more accurate and detailed information, UAV visual inevitably from2D to3Din applications.3D Reconstruction of tree makes more research significance due to itscomplex structure.In this paper, we put forward a complete crown3D reconstruction algorithmbased on the UAV aerial images taken from a certain mountain area in Shanxi.Considering the aerial image contains limited canopy texture and contour information,we proposed a feature extract method and match method based on Local AreaCorrelation Coefficient (LACC), in order to full extract the feature points that canreflect the structure of the target and match the feature points accurately. Furthermore,aerial imaging model has some peculiarity due to the tilting and jitter caused byshaking and tilts of UAV, we estimated inside and outside parameters targeted to thismodel and calculate the3D information of the feature points based on the principle ofbinocular stereo vision. Tree has its unique shape characteristics determined by itsown branches structure, so we proposed a tree modeling method based on L system toSimulate tree growth. Do the experiments on the canopy image cut from the UAVaerial image manually, we found that this method can fully extract the feature pointsof canopy structure and matching feature points accurately, and finally reconstruct theapproximate3D model of the tree from the limited information of aerial images.The feature points extraction and matching method in this paper fully consideredthe characteristics of aerial images and the target to reconstruction, provided somereference significance for the future works; while the modeling method, consideringthe canopy structure characteristics, make full use of the existing information, on ahigher difficulty problem achieves a good effect, is a good tree modeling ideas.