Pixel-based versus object-based identification of scenic resources using Gaofen-2 images: A case study of Yesanpo National Park

Abstract

Scenic resources can serve as symbols of a region’s natural resources and culture and are often the stimulus for the development of national parks. Thus, careful scientific planning and effective management based on the identification and evaluation of scenic resources are key for the sustainable development of national parks. In this study, one object-oriented and three pixel-based (maximum likelihood classification, neural network, and support vector machine) classification methods were applied to identify scenic resources in Yesanpo National Park using high-resolution Gaofen-2 images. The classification accuracy of these scenic resources was evaluated through systematic sampling, which improved the objectivity and accuracy of the classification precision evaluation. All methods met the precision requirements of scenic resource identification, and the accuracy of object-oriented classification was the highest. The application scope of the different methods varies, and suitability can be determined according to the needs of scenic resource recognition. Collectively, this study has proposed an effective and practical method for the identification of scenic resources within Yesanpo National Park, which is of significance for its future planning and management. Moreover, this strategy can be applied by other national park planners to select areas for tourism development, formulate sustainable development strategies, and provide technical support and decision-making guidance for national park planning and management.