| Leaf Area Index(LAI)is an important biophysical parameter and an important parameter in various ecological models,productivity models and carbon cycle studies,etc.It has a strong correlation with crop yield and biomass,and is an important indicator to reflect the growth condition of crop groups.Combining computer technology and agricultural production is an important research direction of virtual reality technology and an important research tool.Three-dimensional reconstruction of agricultural and forestry crops is important for studying the biological characteristics of crops,canopy light distribution,etc.At present,UAV low-altitude remote sensing cost-effective remote sensing monitoring system has achieved more results in agricultural research,but there are fewer related studies using UAV visible light images applied to crop LAI estimation.The study combined UAV point cloud modeling,3D image processing technology and hemispheric camera system to demonstrate the feasibility of UAV aerial imagery for crop LAI estimation.Aerial images of rice at the tillering and gestation stages and tobacco at the root-extension and vigorous stages were collected,and four flight heights of 10 m,20 m,30 m and 50 m were set for each growth period of the two crops,respectively.In order to simplify the computation process,the 3D models of the two crops at each flight altitude were pre-processed with noise reduction and alignment,and then the LAI of rice and tobacco were obtained based on the models combined with the hemispheric photogrammetry principle.Nine rice samples were collected at each of the four heights.In the tobacco model,the LAI was estimated by averaging the nine plants photographed by the fisheye lens camera,but the LAI was averaged by combining the top-view and bottom-view methods during the peak season of tobacco because of the serious mutual shading of the leaves;accordingly,in the canopy structure model,in addition to obtaining the porosity of the plant from the upper viewpoint,the point cloud of the target plant was also rotated by 180° to obtain the LAI from the lower viewpoint.Therefore,18 samples were collected at each of the four heights during the peak season of tobacco.The LAI of the whole plot was calculated by using the single angle method and Miller’s integral equation method,and finally the estimation range was increased from "point" to "surface".The main research contents and results are as follows:(1)The research experimented to obtain canopy porosity by using18°,10°,5°,2.5°,1.25° and 1° as the viewpoint division scale,among the six division methods,the porosity variation under 18° and10° is more fluctuating,and the porosity variation under 5° is more consistent with 2.5°,1.25° and 1°.Taking into account the efficiency and accuracy of data processing,5°,i.e.,18 concentric rings to divide the plants,was chosen as the best canopy division scale.(2)The size of LAI and its spatial distribution are closely related to the number of point clouds and their distribution in space,and the altitude of UAV flight largely determines the density of point clouds and their distribution.Among the four flight altitudes,the 3D models based on 10 m and 20 m flight altitudes can get good results in estimating LAI.30 m flight altitude has lower accuracy,but it requires less data and can be adopted when the accuracy requirement is not high;50 m flight altitude has the lowest accuracy in estimating LAI.(3)In the two growth stages of rice,the Miller integral formula method can achieve better estimation accuracy.In the tobacco root extension stage,the single angle method has better estimation results,with RE,RMSE and R2 at 10 m height of 3.44,0.14 and 0.884,respectively;while the Miller integral formula method has higher estimation accuracy in the vigorous stage.The single-angle method is simpler,but it considers the voidiness at a single viewpoint,and the Miller integral formula method applies the porosity on the whole canopy to the calculation of LAI,so its measurement results are better,and the estimation accuracy decreases when the flight height is higher.In summary,the study concludes that the point cloud is constructed based on UAV aerial images to generate a crop canopy structure model,which can be applied to the extraction of crop LAI,and the crop LAI on the plot unit can be directly obtained,and the technical method of the study adopts a non-contact approach to the crop for data acquisition,with low equipment cost and technical cost,which has good scalability. |
