| As an important food crop,rice plays an important role in food security,and the monitoring of rice growth is very important.As an agricultural parameter that reflects the accumulation of dry matter in rice,AGB(Above-ground Biomass)is of great significance for the final yield estimation of rice.Remote sensing is widely used as a non-contact and efficient monitoring method.The empirical model constructed with vegetation index can be used to effectively estimate the AGB of rice.However,the vegetation index has some limitations in the estimation of AGB.Therefore,it becomes necessary to compensate for the shortcomings of the vegetation index through some other characteristic parameters.In this paper,the double-single-cropping rice paddy area in Central China in Ezhou City,Hubei Province is the experimental area,and the ASD(Analytical Spectral Devices)hyperspectrometer of the ground platform,the RGB sensor based on UAV platform and the multi-band image sensor are used to obtain the experimental data.These data are used to calculate vegetation index and explore the changes of rice canopy reflectance.And the correlation of abundance,image texture and canopy height to rice biomass are compared and analyzed,and respectively combined with the vegetation index to estimate the AGB of rice.The accuracy of regression is used to determine whether the three characteristic parameters can improve the AGB estimation ability of vegetation index.Finally came to the following conclusions:(1)In the rice canopy spectrum obtained by the UAV platform,the red and blue bands of visible light have small differences over time in various fields,but the differences of different fields in the infrared band are large.Subsequently,the calculated vegetation index is used to estimate the AGB.The saturation of vegetation indices related to the red edge band is weaker and the estimation accuracy is higher.Other vegetation indices are obviously saturated,and the accuracy of some vegetation indexes is low.For example,the verification set R~2 of NDVI is only 0.216.(2)The three characteristic parameters are used in this paper:the unmixed abundance map of the mixed pixels,the gray level co-occurrence matrix in the texture and the height of the canopy,which represent the three types of information respectively.The abundance represents the proportion of the corresponding features in a pixel,and for rice fields with a single ground feature,it can indirectly reflect the coverage of the paddy rice.The texture information reflects the horizontal structure of the rice plants to a certain extent through the gray-level co-occurrence matrix.The height of the canopy is a good complement to the structure of the rice in the vertical direction.Through the correlation to biomass,it is found that the abundance of rice,the mean value of 700nm,850nm and 900nm in the gray-level symbiosis matrix,and the canopy height,all show high correlation to the AGB of rice.(3)By combining the three characteristic parameters with the vegetation index and estimating the biomass,we found that the improvement of the saturation phenomenon of the vegetation index only performed poorly on the texture information,and both the abundance of rice and the height of the canopy can improve well.Among the texture characteristics,the mean value of the gray-level co-occurrence matrix at 700nm improves the estimation accuracy of NDVI,but it reduces the estimation accuracy of other vegetation indices.The mean values of 850nm and 900nm have improved the accuracy of vegetation index estimation of biomass,but the overall effect is not significant. |
