王利民, 刘 佳, 杨玲波, 陈仲新, 王小龙, 欧阳斌. 基于无人机影像的农情遥感监测应用[J]. 农业工程学报, 2013, 29(18): 136-145. DOI: 10.3969/j.issn.1002-6819.2013.18.017
    引用本文: 王利民, 刘 佳, 杨玲波, 陈仲新, 王小龙, 欧阳斌. 基于无人机影像的农情遥感监测应用[J]. 农业工程学报, 2013, 29(18): 136-145. DOI: 10.3969/j.issn.1002-6819.2013.18.017
    Wang Limin, Liu Jia, Yang Lingbo, Chen Zhongxin, Wang Xiaolong, Ouyang Bin. Applications of unmanned aerial vehicle images on agricultural remote sensing monitoring[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2013, 29(18): 136-145. DOI: 10.3969/j.issn.1002-6819.2013.18.017
    Citation: Wang Limin, Liu Jia, Yang Lingbo, Chen Zhongxin, Wang Xiaolong, Ouyang Bin. Applications of unmanned aerial vehicle images on agricultural remote sensing monitoring[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2013, 29(18): 136-145. DOI: 10.3969/j.issn.1002-6819.2013.18.017

    基于无人机影像的农情遥感监测应用

    Applications of unmanned aerial vehicle images on agricultural remote sensing monitoring

    • 摘要: 该文以中国农业科学院(万庄)农业高新技术产业园及周边地区4.2×3.1 km的范围为研究区域,利用无人机搭载RICOH GXR A12型相机进行了航拍试验,主要测试了定位定向系统(positioning and orientation system,POS)数据辅助下光束法区域网平差方法平面定位及面积测量精度,以及无人机影像的作物面积识别精度。结果表明,在无控制点约束条件下,直接采用POS数据进行光束法区域网平差后,以中误差表示的平面定位精度为X轴方向(东西方向)中误差为2.29 m,Y轴方向(南北方向)中误差为2.78 m,整体平面中误差3.61 m;采用3阶一般多项式模型进行几何精校正,X轴方向中误差为1.59 m,Y轴方向中误差为1.8965 m,整体平面中误差为2.32 m,符合《数字航空摄影测量空中三角测量规范》中对1∶10 000平地的平面精度要求,能够满足农作物面积遥感监测中作物面积调查定位精度的要求;采用监督分类和面向对象分类2种方法,对面积评价区域种植的春玉米、夏玉米、苜蓿和裸土4种地物类型进行分类,以差分GPS调查结果为评价标准,4种作物总体识别精度分别达到了88.2%(监督分类)和92.0%(面向对象分类),单独分类精度分别为88.9%、86.7%、93.0%、86.6%和90.35%、92.61%、94.93%、93.30%。研究结果说明了无人机遥感影像获取小范围、样方式分布的作物影像方面具有广泛的应用前景,推广后能够满足全国农作物地面样方对高空间分辨率影像的需求,可以部分替代现有人工GPS测量的作业方式。

       

      Abstract: Abstract: By taking Agricultural High-tech Industrial Park of Chinese Academy of Agricultural Sciences (Wan Zhuang) and its peripheral regions with a total area of 4.2 × 3.1 km as the study area, this paper carried out an aerial photogrammetry experiment by using the RICOH GXR A12 camera carried on an unmanned aerial vehicle (UAV), and the experiment mainly tested the precisions of planar positioning under a POS (positioning and orientation system) supported bundle block adjustment method and of area measurement, as well as the precision of the crop area identification of an UAV orthophoto map obtained from an aerial triangulation correction. We use an unmanned aerial vehicle (UAV) to obtain 690 images which covered the whole study area. After a series of processes such as image screen, POS-supported aerial triangulation correction, digital elevation model making, image fusion, and digital differential rectification, we have obtained the ortho-photo map of the whole study area. Since the deployment of high precision ground control point wastes time and energy, POS-supported aerial triangulation employs a non-control point model. Therefore, its absolute positioning precision may be affected by the error of the GPS carried on an UAV. In order to eliminate this error, the project team used a high precision wordview image to rectify the ortho-photo map. In this way, we could improve the image positioning precision, and meanwhile unify the study sample areas with the overall larger scope image coordinate system, so as to provide high precision samples for large-scale agriculture remote sensing statistics and monitoring. The result shows that, under the condition of no control point and after direct POS data bundle block adjustment, the mean square error of plane positioning precision of the X axis direction is 2.29 m, Y direction is 2.78 m, and overall plane error is 3.61 m. If a three order general polynomial model is adopted to conduct a geometric precision correction, then the mean square error of the X axis direction is 1.59 m, the Y direction is 1.8965 m, and the mean square error of the overall plane is 2.32 m. The above figures conform to the 1:10 000 ground plane precision requirements specified in the 'Standard for Aerotriangulation of Digital Aerophotogrammetry' and can meet the positioning precision requirements of a crop area survey in remote sensing monitoring. After obtaining the ortho-photo map, the four ground objects in the area evaluation areas of spring corn, summer corn, alfalfa, and bare soil were classified by employing two methods of supervised classification and object-oriented classification. By taking the differential GPS survey results as the evaluation criteria, the overall precisions of the four crops reached 88.2% (supervised classification) and 92.0% (object-oriented classification) respectively. The separate classification precisions of the two classification methods of the four ground objects were 88.9%, 86.7%, 93.0%, 86.6%, and 90.35%, as well as 90.35%, 92.61%, 94.93%, and 93.30% respectively. The result showed that remote sensing images of unmanned aerial vehicle (UAV), by acquiring small scale and quadrat sampled crop images, have a prospect of wide application. After promotion, it can meet the demands of nationwide crop ground sampling on high spatial resolution images, and can partially replace the operation model of GPS measurement.

       

    /

    返回文章
    返回