摘要
中国芒果种植面积居世界第二,并有持续增长的趋势。国内外利用遥感手段提取农作物的相关研究较多,但有关芒果林遥感提取的研究仍较少。本研究基于2016年12月的高分辨率卫星SPOT-6数据,结合植被覆盖度(FVC)和坡度(SLOPE)因子,利用面向对象分类方法对芒果林地信息进行提取,结果表明,利用FVC和SLOPE参与分割的面向对象分类方法,提取芒果林地的生产者精度达92.81%,用户精度达97.19%。该方法相比于最大似然法分类以及FVC和SLOPE未参与分割的面向对象分类,提取芒果林地的生产者精度分别提高了19.28%和3.62%;用户精度分别提高了8.62%和3.86%。本研究可为果园用地的遥感识别与信息提取有效的方法借鉴。
China has the second largest mango acreage in the word, and the planting area of mangos has followed a trend of sustained growth. Remote sensing has proved very useful in collecting information about the area of cereals and forest, but has rarely been applied to mango orchards. To utilize remote sensing in mango data collection, we attempted to estimate the acreage of mango orchards in China. Data from SPOT-6 and ASTER GDEM covering the study area was used. The boundary mask and sample points were used for acreage estimation. Vegetation coverage （FVC） and SLOPE factors were used to extract information about mango orchards using the objectoriented method. We found that mango orchard producer＇s accuracy obtained by object-oriented classification and FVC and SLOPE data segmentation is as high as 92.81% ; user accuracy is as high as 97.19%. Compared with maximum-likelihood classification and object-oriented classification without FVC and SLOPE data segmentation, object-oriented classification with FVC and SLOPE data segmentation has a higher accuracy for mango orchard extraction： producer accuracy increased by 19.28% and 3.62%, and user accuracy increased by 8.62% and 3.86%, respectively. Therefore, object-oriented classification with auxiliary data improves the accuracy of mango orchard extraction and this method provides an effective technical reference for other artificial orchards.
出处
《资源科学》
CSSCI
CSCD
北大核心
2017年第8期1584-1591,共8页
Resources Science
基金
海南省自然科学基金项目(2016CXTD015)
海南省应用技术研发与示范推广专项(ZDXM2015102)
海南省重大科技计划项目(ZDKJ2016021-02)
关键词
芒果林地
面向对象分类
植被覆盖度
数字高程模型
遥感提取
mango orchards
object- oriented classification
vegetation coverage
digital elevation model
remote sensing extraction
