| Winter wheat height is one of the key factors in monitoring crop growth status and growth rate,and plays an important role in predicting crop yield and optimizing agricultural production management.Compared with large-scale satellite remote sensing images and high-cost Light Detection and Ranging(Li DAR)point clouds,The Structure from Motion(Sf M)algorithm based on digital images can generate a 3D point cloud of the target area at a low cost to accurately extract crop height.UAV Sf M point cloud is widely used in forest detection,urban planning,and many other fields due to its light and flexible UAV platform and the data acquisition advantages of the Sf M algorithm.However,the Sf M algorithm only captures the surface points of the target object,unable to penetrate the dense crop canopy to generate ground points.When winter wheat reaches the stem extension stage,the ground is gradually blocked by leaves,and the absence of ground points in the UAV Sf M point cloud becomes more and more obvious,which makes the ground fluctuation seriously affect the results of crop height extraction.Therefore,this study was carried out to address the above problems.In this thesis,the background and significance of winter wheat height extraction based on UAV Sf M point cloud data are described.Meanwhile,the main point cloud processing algorithms at home and abroad are sorted out,and some important and innovative filtering algorithms are analyzed.At the same time,the organizational characteristics of UAV Sf M point cloud data and some common point cloud preprocessing processes are summarized.The key problems of the current methods of extracting winter wheat height through point cloud data are summarized,and two methods of extracting winter wheat height are proposed.The first method proposed in this thesis is to directly calculate the height of winter wheat by fitting the missing ground points to reduce the interference of unrelated canopy points and abnormal outliers on height calculation.A canopy slice filter is designed to extract the effective information in the winter wheat canopy.The Random Sample Consensus(RANSAC)algorithm was introduced to extract the real ground points from the filtered effective point cloud,and the elevation information of the missing ground points was fitted by reverse distance weighting according to the real ground points,to reconstruct the missing ground fluctuation and estimate the winter wheat height.The results showed that R~2,RMSE,and MAE were 0.90,3.6 cm,and 2.5 cm respectively.The results showed that the height of winter wheat at the tillering stage and stem extension stage was detected accurately.In addition,the effects of the canopy slice filter and fitted ground points on the height estimation accuracy are compared and analyzed respectively.The results show that the canopy slice filter effectively optimizes the extraction efficiency of real ground points,the missing ground points can approximately restore the topographic relief,and the height estimation accuracy is improved.When winter wheat grows to heading stage and later,the actual ground points from the original point cloud are too few and concentrated near the road.To solve this problem,the second method proposed in this thesis is to obtain the height of winter wheat indirectly through canopy fluctuation model:the winter wheat canopy surface model of the original point cloud is extracted by using the moving window,and the canopy fluctuation model is obtained through thinning.The canopy fluctuation model is lowered and spliced with a small number of real ground points,and the point cloud continuity of the spliced ground fluctuation model is corrected to approximate restore the ground fluctuation in the experimental area.From the perspective of the whole growth cycle of winter wheat,the height estimation results of this method were always close to the measured height,and the R~2,RMSE,and MAE were 0.94,4.0 cm,and 3.5 cm respectively.The height estimation accuracy of this method was lower than that of the direct method when they were at the tillering stage and stem extension stage,but it was higher than that of the direct method at the booting stage and heading stage.Therefore,the indirect method has wider applicability and is more conducive to the realization of long-period monitoring of winter wheat height. |
