Rapeseed LAI Estimation Based On UAV Multisource Remote Sensing Data

Oilseed was the third most important worldwide oil crop,behind soybean and oil palm by 2016.Moreover,it is one of the major oil crops cultivated in China.The seedling stage accounts for half of the whole growth period of rapeseed,which is the key period for rapeseed nutrient accumulation.Therefore,it is important for farmers to understand the growth status of rapeseed in the seedling stage and take relative measures accordingly.LAI(Leaf area index)is usually defined as the sum of the leaf area per unit of land area,which is not only closely related to vegetation transpiration,photosynthetically active radiation,but also often serves as an important basis for evaluating growth.Traditional LAI acquisition mostly relies on manual or satellite remote sensing methods,which is difficult to apply to small-field crop LAI monitoring.The development of UAV remote sensing technology and image processing algorithms have provided strong support for the LAI monitoring of field crops.At present,there are many types of UAV based sensors,but there are differences in their price,operational complexity,and actual LAI monitoring effect.Therefore,in order to explore the practical effect of different UAV based sensors in the monitoring of rapeseed LAI at seedling stage,this study used UAV to load consumergrade sensors,multispectral sensors,and hyperspectral sensors for image acquisition.Based on the acquired data,the mixed pixel vegetation index(VImix),rapeseed pixel vegetation index(VIgreen),and comprehensive index(VIgreen * plant height(PH))were calculated.On this basis,this study combined with the field measured LAI,established the LAI inversion model of the above index,and verified the accuracy of the estimated model.Based on the above research content,the main results of this article are as follows:(1)When using the VImix for LAI prediction,the LAI prediction effect based on the NIR(near-infrared)and red edge(RE)index of each sensor is significantly better than that of the RGB vegetation index,and the performance of the hyperspectral sensor is significantly better than that of modified consumer-grade sensing and multispectral sensor.In addition,the optimal band combination based on hyperspectral sensors,multispectral sensors,and modified consumer were 607.026 nm and 761.529 nm,RE and NIR,NIRNP788(788nm long pass band)and RIB(Red band from infrared block filter),respectively.(2)Due to the influence of the exposure and the sensor's spectral response capability,the imaging quality of the broadband image of the modified consumer sensor is significantly better than that of the narrow-band image.Therefore,the performance of broadband VIs was better than that of narrowband VIs.This study proposes a near-infrared modification method for consumer-grade sensors.The principle of this solution is: on the premise of ensuring that the spectral signal passing through the filter is sensitive to LAI,selecting a filter with a wider band width for near-infrared modification.(3)The LAI prediction effect based on the VIgreen of each sensor has basically improved compared to the VImix.Among them,the increasing trend based on consumer sensors and multispectral sensors is relatively obvious,and the improvement based on hyperspectral sensors is not obvious.(4)Because there is a certain relationship between plant height and the number of lower leaves in rapeseed at seedling stage,the VIgreen * PH added with PH information is significantly better than VIgreen in predicting LAI.Hyperspectral sensors cannot obtain DSM(Digital surface model)data,while consumer-grade sensors and multispectral sensors can obtain high-precision DSM data,so as to obtain accurate plant height information of rapeseed at seedling stage.Therefore,comprehensively considering the spectral signal capture capabilities of different sensors,the accuracy of imaging texture,and the accuracy of plant height acquisition,this study finds that the LAI prediction accuracy based on the VIgreen * PH of consumer sensors is close to the highest accuracy based on multispectral and hyperspectral sensors.In summary,the application of low-altitude UAV remote sensing technology has great potential in seedling rapeseed LAI prediction.UAV-based modified consumer-grade sensors with its low cost,easy operation,and high efficiency can be applied to the LAI monitoring of rapeseed or other crops,providing a reference for agricultural production.