农用无人机多传感器遥感辅助小麦育种信息获取

摘要

为实现小麦育种过程中大规模育种材料表型信息快速高通量获取，该文分别从无人机平台优选、农情信息采集传感器集成及数据处理与解析等方面开展研究，研发了一套农业多载荷无人机遥感辅助小麦育种信息获取系统。该系统基于多旋翼无人机平台，并集成高清数码相机、多光谱仪、热像仪等多载荷传感器，提出了无地面控制点条件下的无人机遥感数据几何精校正模型，实现多载荷遥感数据几何校正。该系统操控简便，适合农田复杂环境条件作业，能够高通量获取作物倒伏面积、叶面积指数、产量及冠层温度等育种关键表型参量，为研究小麦育种基因型与表型关联规律提供辅助支持。%To realize rapid acquisition of massive phenotypic information of wheat breeding material, the studies on UAV (unmanned aerial vehicle) platform selection, sensor integration, and remote sensing data processing and analyses were carried out respectively, and a set of multi-load agricultural UAV based remote sensing system for assisting crop breeding information acquisition was developed. The speed and height of multi-rotor UAV could be controlled easily, even at low altitude. Different kinds of sensors such as digital camera, multi-spectral camera and infrared thermal imager, could be loaded on the UAV at the same time. The above characteristics make multi-rotor UAV most suitable to acquire different kinds of farmland spatial information at different scales readily. The crop lodging area could be obtained according to the generated ortho-image based on the high-definition digital images. The crop growth status and crop coverage could be estimated through multispectral images. The canopy temperature, as an important index related to crop growth, could be rapidly acquired by a thermal infrared sensor. The developed system was used in the breeding experiments in the breeding base of Academy of Agricultural Sciences in Lixiahe, Jiangsu from March to June in 2014. In order to acquire orthophotos, canopy spectrums and temperature of wheat breeding area, the Canon PowerShot G16 digital camera, the Tetracam ADC Lite multi spectral camera and the Optris PI thermal imager were loaded on the multi-rotor UAV. The position and altitude parameters of UAV were acquired by GPS (global positioning system) and IMU (inertial measurement unit) sensor simultaneously. The ASD FieldSpec Pro spectrometer collected reflectance data of black and white calibration cloth, cement and water synchronously. The data were subsequently used in multispectral camera radiometric calibration. The data were also tested by the FLIR SC620 thermal imager and the SPOT THERMOMETER HT-11D infrared radiometer for validating the acquired temperature data. At the same time, leaf area index (LAI) data of each plot were collected, and their accurate positions were recorded by hand-held differential GPS (centimeter level). The boundary extraction of breeding subarea and the lodging area estimation were conducted by the image recognition and artificial discrimination in the study. The lodging area estimated was up to 49.88 m2. Multispectral reflectance images were generated by strict radiometric calibration after accurate geometric correction. Vegetation indices, such as normalized difference vegetation index (NDVI), optimization soil adjusted vegetation index (OSAVI) and nitrogen reflectance index (NRI), were computed, respectively. The results showed that NDVI had a strong correlation with LAI (R2=0.48, RMSE=0.27,n=8). The wheat yield forecast was carried out by the nitrogen fertilization optimization algorithm in which INSEY (in-season estimate of yield) index was calculated by the NDVI and local weather data. The yield prediction model was established(R2=0.722, RMSE=0.45,n=25). In this study, the temperature of sky and ground observed were used in the thermal imager temperature correction. Quick mosaic was done to temperature image after correction and then canopy temperature data were extracted. Wheat canopy temperature acquired synchronously by infrared radiometer was combined in the validation with the accuracy (R2=0.84 and RMSE=1.77,n=14).