Research On Key Parameter Extraction Method Of Farm Crops Growing Based On Low Altitude Airborne Sensor Technology

In Xinjiang's agricultural production,the most closely related to precision agriculture is the monitoring of crop growth.Among all the parameters,the height of the canopy,the LAI index and the moisture content in the soil are the key parameters.The traditional approach to monitoring the growth is to rely on manual surveys to accomplish,subjective and inefficient.Precision agriculture on the one hand requires a shorter period of time,on the other hand also need more accurate results,satellite remote sensing sometimes can not meet.Low-altitude unmanned aerial vehicle system with higher spatial resolution and time resolution,higher mobility,is very suitable for farmland monitoring work carried out.Based on this practical need,this study uses a variety of low-altitude airborne sensors to obtain the key growth parameters of farmland crops.The specific contents and achievements are as follows:(1)Classification of farmland using visible light camera dataThe color field conversion and gray level co-occurrence matrix texture filtering method,color index method and object-oriented information extraction method are used to classify the visible light camera data,and good results are achieved.However,according to the classification accuracy,the accuracy of the object-oriented classification method is higher than that of the color space conversion and the graylevel co-occurrence matrix texture filtering method,which is the most effective and accurate classification method.(2)Multi-spectral data of farmland LAI inversionBased on the four different vegetation indices calculated by 4-channel multi-spectral image data,a linear regression model and quadratic polynomial nonlinear regression model were constructed respectively.Compared with the measured LAI data,it is found that the accuracy of the REVI and RVI indices is 97.58% and 96.09%,respectively.(3)Lidar data inversion of crop height and LAI index(NDSM)of the cotton in the farmland was obtained by using the digital elevation model(DEM)and the digital surface model(DSM).The maximum value of the absolute error was only 0.1259 m The In addition,the farmland crop coverage was obtained,and the LAI result of the region was obtained according to the LAI formula.The coefficient of determination was 0.8624.(4)Multi-spectral data of soil moisture inversionThe soil moisture under no soil cover,medium coverage and high vegetation coverage was retrieved by soil thermal inertia method,PDI,SPDI,MPDI and MMPDI indices before and after planting in cotton field.The results of the simulation of the thermal inertia,NDWI index and the measured data were very good,R2 reached 0.9132.In the cotton field with moderate vegetation coverage,the inversion accuracy of SPDI index was the highest.Under the condition of vegetation coverage,the MMPDI index was the best method,R2 was 0.6187.(5)Longitudinal parameter information extraction comprehensive verificationA comprehensive experiment was carried out to collect the image and ground verification data of the same location in the same period,and to analyze and verify the reliability and feasibility of the best method in the previous conclusion.The results showed that the relative error value of farmland classification was relatively small and the precision was high.The values of LAI were 90.8256% and 89.4944% respectively by using GNDVI and RVI index.The results of soil moisture estimation using MMPDI index were better than those of PDI results showed that the plant height of the cotton obtained with the laser radar was 89.358% and the EA value of the LAI index was 94.7199%.