Abstract: Estuary wetland is rich in biodiversity and highly heterogeneous in landscape pattern. Because of the complexity of estuary wetland landscape, traditional classification methods based on single remote sensing images cannot obtain high precision classification results. In this study, multitemporal UAV (unmanned aerial vehicle) remote sensing images were used to optimize the automatic classification results of estuary wetland landscape. The Pingqiao River estuary wetland in the upper reaches of Tianmu Lake was selected as a study case with four season UAV images. The classification method of object based and decision tree was optimized to classify the images from different combinations of four seasons. The results showed that the adoption of multitemporal UAV images significantly improved the classification precision. The more time of UAV images involved in classification, precision will be better. Spring was the most suitable season for landscape classification among the four seasons, with the overall accuracy of 62.7% and Kappa coefficient of 0.59. When the images from all four seasons were used in classification, the overall accuracy increased to 91.7% and the Kappa coefficient was 0.90. The greater the difference in the multitemporal spectral characteristics in classification, the more obvious the classification precision enhancement was. Our results provide technical support for classification of estuary wetland landscape, and put forward a new idea for wetland landscape classification by using visible-light UAV remote sensing images.

Key words: unmanned aerial vehicle (UAV), estuary wetland, multi-temporal remote sensing image, object-based, landscape classification.