CAJ | 학술논문

　　依据江苏盐城国家珍禽自然保护区淤泥质潮滩湿地影像特征，快速提取高精度潮滩湿地地物信息对湿地生态保护具有重要的意义。本研究以2010年TM影像为数据源，针对海滨湿地植物覆被类型复杂，以及湿地植物类型之间的生态交错带信息难以识别等问题，综合运用植被NDVI指数、波段反射率特征、环境特征和生态条件，逐级分层分类及人工选取阈值等方法，较好地解决了淤泥质潮滩湿地分类问题。结果表明，与同一时期的监督分类相比较，在识别植被交错带植被覆被类型和零星分布的植被斑块的类型方面更具优势，分类精度有明显提高。通过ROI训练区，选取了3126个包括所有类型的样本进行精度检验，分类精度达到95.87%。该方法弥补了单一分类方法的不足，对快速、高精度地提取淤泥质潮滩地物类型具有重要的参考价值和实践意义。
　　의거강소염성국가진금자연보호구어니질조탄습지영상특정，쾌속제취고정도조탄습지지물신식대습지생태보호구유중요적의의。본연구이2010년TM영상위수거원，침대해빈습지식물복피류형복잡，이급습지식물류형지간적생태교착대신식난이식별등문제，종합운용식피NDVI지수、파단반사솔특정、배경특정화생태조건，축급분층분류급인공선취역치등방법，교호지해결료어니질조탄습지분류문제。결과표명，여동일시기적감독분류상비교，재식별식피교착대식피복피류형화령성분포적식피반괴적류형방면경구우세，분류정도유명현제고。통과ROI훈련구，선취료3126개포괄소유류형적양본진행정도검험，분류정도체도95.87%。해방법미보료단일분류방법적불족，대쾌속、고정도지제취어니질조탄지물류형구유중요적삼고개치화실천의의。
Remote sensing technology is a kind of effective technical means to obtain information of the tidal muddy flat wetlands. It has important significance to explore the tidal flat wetland’s remote sensing classifica-tion method. This study’s aim is to get high-precision information of features of tidal flat wetlands according to image feature and muddy tidal flat wetlands characteristics of Yancheng National Nature Reserve in Jiangsu Province, China. The TM image of 2010 is taken as the data source. The practical problems are tidal flat wetland vegetation types’complexity and the difficulty in correctly identifying the ecotone information. Because tidal flat wetland surface features in remote sensing images have a complex spectrum, it is very difficult to effectively extract information through one single method. So, we comprehensively use methods of the NDVI, band reflec-tance spectral characteristics, environmental characteristics and ecological conditions, one level after another to simplify the complex tidal flat wetland classification. Firstly, we chose the best combination of the bands. Sec-ondly, we established vegetation and non-vegetation’s ROI to analyze the band reflectance spectral characteris-tics. The ROI include 1316 Spartina alterniflora Loisel, 633 mudflats, 1253 Suaeda salsa, 1002 Phragmites cam-munis, 525 water body, and 4803 road samples. At last, we established decision tree to reality the practices. After selecting 3216 samples to accuracy inspection and compared with the same area of supervised classification, the results show that classification accuracy is up to 95.87%by ROI training area. It has advantages in identifying the ecotones’vegetation types and vegetation scattered patches. The method makes up for the lack of a single classification and has an important reference value and practical significance in effectively extracting tidal flat surface features.