基于TWDTW的黄河三角洲湿地植被分类研究
Classification of Wetland Vegetation in the Yellow River Delta Based on TWDTW

湿地植被是湿地生态系统中必不可少的一部分，在调节气候、蓄洪防旱等方面发挥了重要作用，对其进行行之有效的监测对于生态保护至关重要。基于时序的遥感影像进行地物分类在植被分类领域具有重要优势，湿地植被生长环境复杂，不同植被之间物候特征并不明确，且滨海湿地地区云雾较多，这些因素限制了时序遥感在湿地植被监测中的应用。基于时间加权的动态时间归整(Time-Weighted Dynamic Time Warping, TWDTW)通过增加时间权重限制实现时序匹配，能够避免植被物候因素的干扰与畸形匹配现象。本文探讨了该算法在黄河三角洲湿地植被分类中的适用性，并将分类结果与传统分类方法进行对比。研究表明，该算法在该区域总体分类精度为97.56%，Kappa系数为0.95。应用TWDTW算法可以有效进行湿地植被分类，能直观地反映湿地植被的空间分布格局，满足湿地生态环境监测、资源调查与管理等方面的需要。
Wetland vegetation is an indispensable part of wetland ecosystems, which plays an important role in climate regulation, flood storage, drought prevention, etc. Effective monitoring of wetland vegetation is essential for ecological conservation. However, the application of time-series remote sensing in wetland vegetation monitoring is limited by the complex growth environment of wetland vegetation, the unclear physical characteristics of different vegetation, and the high cloudiness of coastal wetland areas. By increasing the time weight limit to achieve temporal matching, the interference and aberrant matching of vegetation phenology can be avoided by TWDTW (Time- Weighted Dynamic Time Warping). This paper explores the applicability of this algorithm to the classification of wetland vegetation in the Yellow River Delta and compares the classification results with traditional classification methods. The study shows that the overall classification accuracy of the algorithm is 97.56% and the Kappa coefficient is 0.95. The application of the TWDTW algorithm can effectively classify wetland vegetation and reflect the spatial distribution pattern of wetland vegetation. In addition, the result can meet the needs of wetland ecological environment monitoring, resource survey, and management etc.