Abstract: Crop height is one of the important indicators in the research of farmland,ecology and other fields. Accurate crop growth monitoring and yield prediction requires the accurate data of crop height and its category. Compared with the traditional remote sensing detection technology,LiDAR,by the advantages of high mobility, easy to carry,strong handling characteristics,has a better remote sensing technology,with which,the non-contact measurement can be realized and massive accurate cloud data can be accessed. In this way,the canopy height can be obtained by obtaining accurate vegetation canopy structure information. In this paper,the cotton field of Oasis farmland is taken as the research object,and the preprocessing process of noise eliminating of the data,data filtering and point cloud dilution was carried out. In the LiDAR point cloud data,the points from the canopy and the ground are distinguished respectively,and the vegetation vertical structure information is obtained. Based on the ground and non - ground points,the digital terrain model (DEM)and the digital surface model (DSM)were obtained respectively. On this basis,a normalized digital surface model (nDSM)was obtained. The pixel value in the nDSM model is the absolute height of the crop after the removal of the terrain factor. The vegetation in the area was classified by estimating the height of each vegetation type in the study area. According to the field sampling data,the average height of woodland,fresh vegetables,cotton and zucchini is about 12.5,2.125,1.125 and 0.35 m. The distribution profiles of four types of vegetation,including forestland,fresh rice,cotton and zucchini were made. It can be seen from the figure that the height of the four typical vegetation types are obvious different. The overall accuracy of the classification results is 93%,the Kappa coefficient is 0.911,and the classification accuracy is very good. In order to verify the accuracy of the method,the correlation between the measured height data and the calculated vegetation height results was analyzed. The decision coefficients R²,absolute error and relative error were selected as evaluation indexes to evaluate the experimental results in the study area. The results showed that the average relative error between vegetation height and vegetation height was -0.038%,1.35% and -4.348%,respectively. This indicates that the vegetation canopy structure parameters extracted from the Li- DAR can be used to estimate the crop height in the farmland area. Therefore,this study provides a way for Li- DAR to extract information about vegetation type differentiation,crop canopy height and small vertical height difference. This study also provides a reliable data basis for the crop growth monitoring and other related research.

Key words: vegetation height estimation, LiDAR, farmland