农业工程学报
農業工程學報
농업공정학보
2014年
9期
148-156
,共9页
康峻%侯学会%牛铮%高帅%贾坤
康峻%侯學會%牛錚%高帥%賈坤
강준%후학회%우쟁%고수%가곤
遥感%植被%分类%决策树%物候%时间序列%MODIS%双Logistic函数拟合
遙感%植被%分類%決策樹%物候%時間序列%MODIS%雙Logistic函數擬閤
요감%식피%분류%결책수%물후%시간서렬%MODIS%쌍Logistic함수의합
remote sensing%vegetation%classification%decision trees%phenology%time series%MODIS%double logistic function fitting
针对目前遥感分类较少利用植被物候特征参数的现状，该研究以内蒙古自治区额尔古纳、根河两市作为试验区，基于2012年根河土地覆盖数据，采用双Logistic 函数拟合的方法对跨度范围为2011年7月下旬至2013年7月下旬的时间序列MODIS EVI 数据进行植被物候参数的分析与提取，并依据物候参数特征构建决策树对试验区土地覆盖进行分类。研究表明，不同植被的物候有较明显的特征，森林、草原与作物的生长季开始时间依次滞后，作物的生长季最短，森林与草原生长季基本持平；利用植被物候特征参数进行决策树分类，作物和森林2种植被类型取得较好分类效果，分类的总体精度达到73.67%，优于该区域 MODIS 土地覆盖产品的总体精度（66.08%）。该研究可为呼伦贝尔地区生态环境评价和农、牧、林业生产活动提供一定的参考。
針對目前遙感分類較少利用植被物候特徵參數的現狀，該研究以內矇古自治區額爾古納、根河兩市作為試驗區，基于2012年根河土地覆蓋數據，採用雙Logistic 函數擬閤的方法對跨度範圍為2011年7月下旬至2013年7月下旬的時間序列MODIS EVI 數據進行植被物候參數的分析與提取，併依據物候參數特徵構建決策樹對試驗區土地覆蓋進行分類。研究錶明，不同植被的物候有較明顯的特徵，森林、草原與作物的生長季開始時間依次滯後，作物的生長季最短，森林與草原生長季基本持平；利用植被物候特徵參數進行決策樹分類，作物和森林2種植被類型取得較好分類效果，分類的總體精度達到73.67%，優于該區域 MODIS 土地覆蓋產品的總體精度（66.08%）。該研究可為呼倫貝爾地區生態環境評價和農、牧、林業生產活動提供一定的參攷。
침대목전요감분류교소이용식피물후특정삼수적현상，해연구이내몽고자치구액이고납、근하량시작위시험구，기우2012년근하토지복개수거，채용쌍Logistic 함수의합적방법대과도범위위2011년7월하순지2013년7월하순적시간서렬MODIS EVI 수거진행식피물후삼수적분석여제취，병의거물후삼수특정구건결책수대시험구토지복개진행분류。연구표명，불동식피적물후유교명현적특정，삼림、초원여작물적생장계개시시간의차체후，작물적생장계최단，삼림여초원생장계기본지평；이용식피물후특정삼수진행결책수분류，작물화삼림2충식피류형취득교호분류효과，분류적총체정도체도73.67%，우우해구역 MODIS 토지복개산품적총체정도（66.08%）。해연구가위호륜패이지구생태배경평개화농、목、임업생산활동제공일정적삼고。
Phenology refers to periodic plant life cycle events influenced by climate and other environmental factors, such as sprouting, flowering, fruiting and leaves falling, etc. Different vegetation types have distinct growth characteristics, and phenology can be a good representative parameter to classify vegetation types. Phonological parametric analysis is mainly used to find out significant changes in specific time points and extract corresponding characteristic VI values, by analyzing a time-series vegetation index, e.g., start of season (SOS), end of season (EOS), length of season (LOS), max of EVI (MOE) and amplitude of EVI (AOE). These key phenology parameters can be used to classify vegetation types. Eerguna and Genhe in Hulunbeier city, Inner Mongolia Autonomous Region were selected as the study area. A double logistic function fitting method was used to smooth the time series MODIS-EVI data. The time range was from the summer of 2011 (DOY=209) to the summer of 2013 (DOY=193), and the total number of images was 46. Then, 100 points of each land cover type (grass, forest, crops, other non-vegetation) were chosen as classification samples. Five key phenological parameters mentioned above were extracted and used to build the decision tree classifier. The overall classification accuracy of the results reached to 73.67%. The results show that vegetation in Hulunbeier northern region had obvious unique features. The season of forest started earliest (145-160 days, DOY, hereinafter), and ended quiet early (250-275 days); the season of grass started slightly later than forest (160-170 days), but the length of season was similar to forest, both were from 90 to 120 days. The season of crops started late and ended early, so the season of crops was short and concentrated, the length of the samples was from 60 to 90 days. The classification achieved better results than MODIS land cover products (66.08%). Except for grass’ user accuracy being a little lower, producer accuracy and user accuracy of the 4 kinds of land cover types reached to 79%. The phonological information extracted in this paper had a high consistency with existing research results;it is shown that monitoring phenology based on time-series EVI data was reliable. This research can provide a reference to ecological environment evaluation and agriculture, animal husbandry, forestry production activities in the Hulunbeier northern region.
