【作者】 李丹丹；

【导师】 周清波；

【作者基本信息】 中国农业科学院 ， 农业资源利用， 2011， 硕士

【摘要】 油菜是我国五大油料作物之首,是重要的食用油源和蛋白质饲料,也是重要的工业原料,其种植面积占全国油料作物总面积的40%以上,产量占全国油料总产量的30%以上,居世界首位。及时了解油菜种植面积,对于加强油菜生产管理,调整农业结构,辅助有关部门制定科学合理的指导政策具有重要意义。利用遥感技术可以获取到不同时间分辨率和不同空间分辨率条件下大范围的地表覆盖信息,信息提取结果可排除人为干扰,具有科学性和客观性。油菜种植面积遥感提取的国内外研究相对匮乏,不能满足实际应用的需要。油菜主要种植区域地块破碎度高,作物插花种植现象严重等都为油菜种植面积信息提取带来了一定难度。本论文对研究的选题、国内外的研究进展、拟采用的研究方法进行了介绍。总结了前人在相关研究中已经取得的研究成果和仍然存在的问题,提出了研究的必要性和可行性。以中高空间分辨率遥感影像数据为主要数据源,针对安徽省研究区进行了油菜种植面积遥感信息提取研究。研究详细的介绍了具体研究内容,研究方法和得出的研究成果。其中,在安徽油菜种植区划中,研究以安徽油菜为对象,结合多年油菜种植面积、油菜生育期、地形因素等主要影响因素进行区域划分,区划结果为安徽省油菜种植面积遥感信息提取中空间抽样样方选取提供参考。在油菜种植面积遥感提取最佳时相研究中,研究以2009年多时相Landsat5-TM数据为例,对油菜各生育期内遥感影像中地表主要植被类型的光谱可分性距离计算,分析研究区内主要植被物候差异和光谱可分性。经J-M距离计算可知,4月上旬为油菜种植面积遥感提取的最佳时相。在油菜种植面积遥感提取最佳波段组合研究中,研究以Landsat5-TM、ASTER、ALOS(AVNIR-2)、SPOT-5遥感数据为例,根据油菜生长期的光谱特征与中高空间分辨率遥感数据的波谱特征,利用最佳指示因子法(OIF)研究油菜种植面积遥感提取最佳波段组合方式。经OIF指数计算,研究区内Landsat-TM、ASTER, ALOS(AVNIR-2)、SPOT-5遥感数据在近红外和可见光波段均可较好的区分油菜与其他地物。在尺度变化对油菜种植面积提取精度影响的研究中,分别采用遥感自动分类法中的非监督分类法、监督分类中的最小距离法、最大似然法和马氏距离法对油菜种植面积进行提取。采用分类精度和分类效率指标两个参数,结合地面GPS调查数据和QUICKBIRD高分辨率遥感影像对分类结果进行精度评价。通过对提取结果混淆矩阵比较,非监督分类结果好于监督分类结果,且效率高。随着遥感数据空间分辨率的增加可提高油菜种植面积遥感信息提取精度。本论文基于中高空间分辨率遥感数据进行了油菜种植面积信息提取,研究结果表明利用中高空间分辨率遥感数据可以高精度提取油菜种植面积。随着遥感技术的高速发展,多传感器多空间分辨率遥感数据大量出现,今后会采用多数据源多方法对遥感提取油菜种植面积进行继续研究。更多还原

【Abstract】 Rape, ranging the first in five oil-bearing crops in China, is the important edible oil source and protein feed, as well as important an industrial material. Taking up over 40% of total cultivated oil-crop area and over 30% of total output of oil plants nationwide, it ranks the first place in the world. Understanding the cultivated area of rape is of significance to enhance its production control, adjust agricultural structure, and make rational guiding policies assisting relevant departments. Through remote sensing technology, the extensive surface covering information can be obtained in different time resolutions and spaces. The man-made interference can be avoided for the information extraction results, and so it is scientific and objective. The Remote Sensing Information Extraction for rape cultivated area is deficient relatively at home and abroad, and can’t satisfy requirements of practical applications. It is a certain difficulty for information extraction of rape planting area due to the fragmentary and dispersive planting lands and the severe crossing cultivation.In this paper, selection of research subjects, research progress at home and abroad, and proposed research methods all are introduced; the results and problems studied before in relevant researches are summed up; and the necessity and feasibility of the research are put forth. In addition, by mid and high space resolution remote-sensing image data as the main data source, the rape-cultivated area remote information extraction is studied aiming at the Anhui research region. The specific contents, methods and achieved results are introduced in detail. The research focuses on the rape in Anhui rape planting area division in combination of major factors such as its planting area, growth period and landforms in the region. The area-division results render reference for space sampling methods in Anhui rape cultivated area remote sensing information extraction.In the research on optimal time phase of the rape cultivated area remote-sensing extraction, the divisible distance of spectra in major vegetation types on the earth’s surface are calculated in remote sensing images in rape growth seasons taking an example of multi-time phase Landsat5-TM data in 2009; the difference of phonological phenomenon spectrum and divisibility for major vegetations within the research region are analyzed. It is known by calculation of J-M distance that the beginning of April is the best time phase for the rape cultivated area remote sensing extraction.In the research on optimal band combination of the rape cultivated area remote-sensing extraction, the optimal indication factor method (OIF) is used to study the optimal band combination modes based on spectral characteristics during rape growth period and wave-spectrum features of mid and high space resolution remote sensing data, taking an example of Landsat5-TM、ASTER、ALOS(AVNIR-2)、SPOT-5 remote-sensing data. Through OIF index calculation, the Landsat5-TM、ASTER、ALOS(AVNIR-2)、SPOT-5 remote-sensing data can better distinguish rape and other crops within near-infrared and visible light bands.In the research on the impact of dimensional changes on rape cultivated area extraction precision, the rape cultivated area is extracted by the non-supervised classification in the remote-sensing auto classification, as well as the minimum distance method, maximum likelihood method and Mahalanobis distance in the supervised classification method respectively. By two indexes, i.e. precision and effectiveness of classification, the classification results are evaluated in precision in combination of ground GPS investigation data and QUICKBIRD high resolution remote sensing image. By comparison of confusion matrix of extraction results, the non-supervised classification result is better than the supervised classification results with high effectiveness. With increase of the remote sensing data space resolution, the rape cultivated area remote sensing information extraction is improved in precision.In the paper, the rape cultivated area information extraction is conducted based on mid and high space resolution remote sensing data. The research results show that the rape cultivated area can be extracted in high precision by mid and high space resolution remote sensing data. With rapid development of remote sensing technology, a great deal of the multi-sensor and multi-space resolution remote sensing data appears. The multi-data source method will be used to continue to further study the remote sensing extraction of rape cultivated area in the future.更多还原