【作者】 刘冰；

【导师】 刘焕军； 康健；

【作者基本信息】 东北农业大学 ， 农业硕士（专业学位）， 2019， 硕士

【摘要】 “十三五”规划和“镰刀弯”政策出台以来,国家就把更多的精力放在农业上,供给侧的结构调整也成为了炙手可热的问题,而如何合理的、科学的调整种植结构也成为了热门研究。就这一研究问题,本文使用2016年23期空间分辨率为250 m的MODIS遥感影像、分辨率为30 m的Landsat 8遥感影像,获取2016年的物候信息,继而对物候信息进行分区分割处理,最终提取黑龙江省2016年的作物种植结构。选取黑龙江典型地区,针对典型地区使用分辨率为10 m的Sentinel-2A高分辨率影像,使用监督分类方法识别作物类型,进一步验证黑龙江省种植结构提取结果的精度,为宏观尺度下调整黑龙江省种植结构提供合理依据。本文主要的研究方法和研究结果如下:(1)本文提出了“物候均一区”的概念,即物候区。将整个黑龙江省分成多个区域,经过评价指数计算后确定物候分区最优尺度为70,最终得到27块物候均一区;然后对每个物候区进行多尺度分割,经分割指数评价后得到多尺度分割的最优尺度为5。(2)经过统计分析得出,纯玉米的种植区阈值是最大的,可达到1.5以上,所以阈值范围在0.8及以上的最小种植结构单元均可分为玉米的播种区;纯水稻种植区的阈值是最小的,范围在-0.4~0,所以当阈值小于0时,都可划定为水稻的单一种植结构单元,因此可以将水稻-玉米的阈值范围设置为0.1~0.4;而水稻-大豆和水稻-大豆-玉米的阈值波动范围较为相似,都集中在0.5左右。(3)2016年黑龙江省作物种植结构提取整体精度达到88.72%以上。从作物种植结构图整体来看,黑龙江省玉米种植面积分布最为广泛,其次是水稻。大部分玉米种植区位于松嫩平原,夹杂着与大豆混作的区域;水稻主要分布在河流沿岸,呈线状分布,且主要集中在三江平原地区;2016年大豆没有单一种植区,通常与玉米,小麦,大米和马铃薯形成混作区。(4)海伦市处于松嫩平原地区,海伦市种植的作物类型丰富,组成的种植结构单元种类多。使用监督分类结果与种植结构提取结果对比可以看出,监督分类结果和种植结构提取的结果相似度高达94.00%,而根据主成分分析的结果进行种植结构提取,整体精度平均可达到88.72%,因此利用MODIS提取种植结构方法可行性高,且快速便捷。(5)依据MODIS-NDVI时间序列数据可以提取大范围区域内农作物物候信息,并进行分割分类处理可以有效的识别大尺度区域中多类农作物空间分布信息,可以足不出户、及时准确获取农作物信息,为农业政策的制定提供合理依据,为农业的可持续发展奠定基础。更多还原

【Abstract】 Since the 13 th five-year Plan and the "sickle bend" policy were introduced,the state hasdevoted more energy to agriculture,and supply-side structural adjustment has also become a hot issue,and how to be reasonable.Scientific adjustment of planting structure has also become a hot research issue,in this paper,23 issues of 2016,the spatial resolution of 250 m MODIS remote sensing image,the resolution of 30 m Landsat 8 remote sensing image,to obtain phenological information in 2016.Then the phenological information was segmented to extract the crop planting structure of Heilongjiang Province in 2016.Select Heilongjiang typical.The main research methods and results are as follows:(1)In this paper,the concept of "phenological-homogeneous zone",that is,the phenological region,is divided into a plurality of regions.After the evaluation index is calculated,the optimal size of the phenological zone is 70,and finally,the waiting area of the 27 blocks is obtained,and then the multi-scale division is performed on each phenological region,And the optimal scale of the multi-scale segmentation is 5 after the segmentation index is evaluated.(2)Through statistical analysis,it is found that the threshold value of pure maize planting area is the largest,which can reach more than 1.5.Therefore,the minimum planting structure units with the threshold range of 0.8 or above can be divided into sowing areas of maize.The threshold value of pure rice planting area is the smallest with a range of-0.4~0,so when the threshold value is less than 0,it can be planned as a single planting structure unit of rice,therefore,the threshold range of rice-maize can be set to 0.1~0.4;The threshold fluctuation ranges of rice-soybean and rice-soybean-maize are similar,both concentrated in about 0.5.(3)The overall precision of crop planting structure in Heilongjiang Province reached 88.72%in 2016.As a whole,the distribution of maize planting area in Heilongjiang Province is the most extensive,followed by rice.Most maize planting areas are located in Songnen Plain,mixed with soybean,rice is mainly distributed along the coast of rivers,and mainly concentrated in Sanjiang Plain.In 2016,soybeans did not have a single planting area and were usually mixed with corn,wheat,rice and potatoes.(4)Hailun City is located in Songnen Plain area.There are many types of crops planted in Hailun City,and there are many kinds of planting structure units in Hailun City.Comparing the results of supervised classification with that of planting structure extraction,we can see that the similarity between supervised classification and planting structure extraction is as high as 96%,while the results of principal component analysis are used to extract planting structure.The overallaccuracy can reach 88.72% on average.Therefore,the method of extracting planting structure by MODIS is feasible,rapid and convenient.(5)According to MODIS-NDVI time series data,crop phenological information can be extracted in a large area,and segmentation and classification processing can effectively identify the spatial distribution information of multiple crops in large-scale areas,and can not go out of the house.Timely and accurate acquisition of crop information provides a reasonable basis for the formulation of agricultural policies and lays a foundation for the sustainable development of agriculture.更多还原