Estimating Wheat Biomass And Tiller Number Based On Terrestrial Laser Scanning

Wheat is one of the world's major food crop and occupies an important position in agricultural production.Monitoring wheat agronomic parameters with rapid,non-destructive and accurate way is very important for understanding the growth details of wheat and predicting yield in advance.Compared with other remote sensing technology,terrestrial laser scanning which can accurately,actively and quickly obtain three-dimensional spatial information of wheat has unparalleled advantages in monitoring wheat growth parameters and structural information.In this study,all point clouds,measured wheat biomass and tiller numbers were obtained simultaneously from three wheat field experiment in different years,test sites,varieties,planting densities and nitrogen levels.The object of this research is to explore the different methods of estimating wheat biomass,comparing the different methods to determine the optimal metric;and predicting tiller number by constructing a new algorithm(the adaptive layering and clustering algorithm)based on the terrestrial laser scanning.This study provides a reference for the acquisition of wheat biomass and tiller number based on terrestrial laser scanning,and provides example for monitoring of other crops using terrestrial laser scanning technology.Firstly,this paper constructs models for monitoring wheat biomass using terrestrial laser scanning.Two types of metrics were obtained by processing point clouds,which are namely the height metrics H95 and the volume metrics volume.The model was built by two kinds of metrics separately with the measured biomass,and ten-crossed check method was used to test the two methods.The results showed that:(1)Constructing the model by using the height metrics H95 with the measured biomass,the coefficient of determination R2 was 0.84 during the whole growth period,and the verification result(R2=0.81,RMSE=1.57 t/ha,RRMSE=32.55%)indicated that the model has good accuracy and adaptability.(2)Constructing the model by using the volume metrics with the measured biomass,the coefficient of determination R2 was 0.86 throughout the entire growth period,and the verification results(R2=0.77,RMSE=1.72 t/ha,RRMSE=35.74%)indicated that the biomass prediction model also has good accuracy and adaptability.(3)Then we compare the two types of metrics.it is more reasonable to use the volume metrics to estimate the wheat biomass.Then,this paper constructed a new algorithm(tiller number extraction algorithm)based on the terrestrial laser scanning to achieve wheat tiller number:1)First,we get any row of wheat point clouds,projecting the row width Y axis onto a plane and reserving the row spacing X and height Z.2)According to the empirical method,we set the cuttting height to 4cm.The number of layers is that the ratio of wheat plant height to cutting height the wheat plant height divided by the slice height,that is,adaptive stratification.Then select two layers as leaf layer and stem layer.The gap between the two layers' cloud points is to determine a cluster of wheat.3)During the previous clustered wheat cluster,a clustering algorithm was introduced to distinguish the shorter-spacing wheat tillers.The results show that:in this paper,by creating a new algorithm to estimate the number of wheat tillers based on the terrestrial laser scanning,the accuracy is R2=0.61,and the RMSE and RRMSE are 106 and 34.53%respectively(the wheat experiment in Rugao in 2017).This algorithm was tested using independent datasets(Xuzhou experiment in 2017 and Rugao experiment in 2018)with the accuracy are R2=0.56,0.65,and the errors are RMSE=143,1 02,RRMSE=26.26%,and 22.84%.