论文部分内容阅读
以评估退耕还林/还草的气候、生态效应为目的,利用多时相SPOT/VEGETATION数据,将监督分类与非监督分类相结合,选定合适的阈值对西北地区植被及作物进行分类。整个分类过程按照紧凑的时间序列进行顺序分类。分类顺序依次是:常青林地、混合林、水体、冬小麦、草地、灌木、春小麦、水稻、湿地、其他杂粮、玉米、马铃薯、沙漠、半沙漠。①常青林:选择1月份资料,常青林的NDVI最大,其阈值一般选择140左右;②混合林:一般选择10月上旬图像。其NDVI较小,只有林、草的NDVI较大,且林地NDVI大于草地,其NDVI值一般在130~140(为便于分类,将NDVI真实值扩大1000倍,下同);③水体:选择12月中旬的图像;④沙漠:选择各种植被最好的时节,如7月下旬、8月上旬的图像。设定阈值小于50;⑤草地和灌木:选择9月下旬图像。选择分区域进行植被分类,并结合图像本身进行分区分类。阈值选取:草地北部60~65,南部80~95;灌木70~75;⑥冬小麦:选择上年11月上、中旬图像。选择阈值将冬小麦和林地共同区域分出,再除去林地,剩下的就是冬小麦的分布区域;⑦湿地、水稻:选择5~6月上旬图像;⑧春小麦:选择4月下旬的图像。这个时期生长的植被包括冬小麦、春小麦以及林地、草地,从此时相的图像中分出所有植被,然后排除冬小麦、林地、草地,即得春小麦种植区域;⑨玉米、马铃薯混合区:图像选择8月中、下旬,宁夏灌区单独分出,灌区可区分出的植物为水稻和玉米,排除水稻区域即为玉米。其他地区为玉米、马铃薯混合区其他旱地作物:主要有油料作物、山区谷类作物。时相选择和玉米、马铃薯混区相同;⑩半沙漠:以上所有植被类型都分类完成以后进行合并,空白的区域认为是半沙漠地区。黄土高原稀疏植被区域,由于植被分布十分稀疏,认为是半沙漠。上述分类结果与统计面积、实际调查结果相比,精度在85%以上。
In order to assess the climatic and ecological effects of returning farmland to forest / grassland, using the multi-temporal SPOT / VEGETATION data, the supervised classification and unsupervised classification are combined to select the appropriate threshold to classify the vegetation and crops in the northwest China. The whole classification process is sorted according to a compact time series. The classification order is: evergreen forest, mixed forest, water body, winter wheat, grassland, shrub, spring wheat, rice, wetland, other grains, corn, potato, desert and semi-desert. ① evergreen forest: select the January data, the evergreen forest NDVI maximum, the threshold generally choose 140 or so; ② mixed forest: the general selection of early October images. Its NDVI is small, only the NDVI of forest and grass is larger, and the NDVI of forest land is larger than that of grassland, and the NDVI value is generally 130-140 (for the convenience of classification, the real value of NDVI is expanded by 1000 times, the same below); ③ Water body: choose 12 Mid-month image; ④ desert: choose the best season for all kinds of vegetation, such as the image of late July and early August. Set the threshold less than 50; ⑤ grass and shrubs: Select the image in late September. Select the sub-region vegetation classification, combined with the image itself partition classification. Threshold selection: 60 ~ 65 in the northern grassland, 80 ~ 95 in the south; 70 ~ 75 shrubs; ⑥ winter wheat: Select the image of the upper and middle of November last year. Select the threshold to separate the winter wheat and woodland common area, and then remove the woodland, and the rest is the distribution area of winter wheat; ⑦ wetland, rice: select the image of early May ~ June; ⑧ spring wheat: select the image of late April. Vegetation in this period includes winter wheat, spring wheat and woodland and grassland. All vegetation is separated from the images of this phase and then winter wheat, woodland and grassland are removed. ⑨ Corn and potato mixed area: Image selection August Middle and late, Ningxia irrigation area separately, the irrigation area can be distinguished from the plants of rice and corn, excluding rice area is corn. Other areas of corn, potato mixed area Other dry land crops: the main oil crops, mountain cereal crops. Phase selection and corn, potato mixed area; ⑩ semi-desert: All the above vegetation types are classified after the completion of the merger, the blank area that is semi-desert. Loess Plateau sparse vegetation area, due to vegetation is very sparsely distributed, that is semi-desert. The above classification results and statistical area, the actual survey results, the accuracy of more than 85%.