What question did this study set out to answer?

This research aims to explore how deep learning-driven urban change detection can support sustainable development in hilly-area cities.

May 26, 2026Scientific Reports0 citationsOpen Access

Research on the application of deep learning-driven urban change detection in sustainable development of hilly-area cities in western China

HWHaiying WangChina West Normal University MWMingzhong WuChina West Normal University

Key Points

This research aims to explore how deep learning-driven urban change detection can support sustainable development in hilly-area cities.
Constructed region-adapted datasets for semantic segmentation and change detection in Nanchong, Sichuan Province.
Applied deep learning models for segmenting buildings and roads, and detecting urban changes using custom datasets.
Conducted comparative experiments to evaluate performance in complex urban environments.
Deep learning-driven change detection effectively addresses challenges in hilly urban settings.
Urban expansion patterns were analyzed, revealing developmental trends and spatial morphology changes.
Demonstrated practical value in optimizing urban structure and improving land-use efficiency.

Abstract

Abstract Urban sustainable development is a critical pathway to overcoming resource constraints in the urbanization processes and achieving synergistic high-quality economic development and ecological security in hilly areas city of western China. This study investigates the application of deep learning-driven urban change detection technology in supporting sustainable urban development, taking Nanchong, Sichuan Province—a typical hilly areas city in western China as the case study area. Three core tasks were conducted. First, region-adapted urban element datasets were constructed, including a building and road semantic segmentation dataset (BRDataNC) and a change detection dataset (CDDataNC), both tailored to the landscape characteristics of hilly urban areas. These datasets provide targeted and reliable support for training and validating deep learning models suitable for medium-resolution remote sensing imagery in hilly regions. Second, deep learning models were applied to conduct semantic segmentation of buildings and roads based on BRDataNC, and further performed urban change detection using CDDataNC. The experiments were carried out entirely on the self-constructed datasets to ensure reasonable evaluation under consistent data characteristics. Comparative experiments demonstrated that deep learning-driven change detection effectively addresses challenges in complex hilly urban environments, such as fragmented landscapes, scattered buildings, and spectrally mixed features. Third, leveraging the change detection outcomes, this study analyzed urban expansion patterns in the research area, uncovering the evolutionary characteristics and potential trends in urban spatial morphology. The findings indicate that deep learning technology offers a robust tool for dynamic urban monitoring and informed decision-making in the context of sustainable urban development in hilly areas cities of western China. This approach exhibits clear practical value for optimizing urban spatial structure, improving land-use efficiency, and supporting coordinated urban development.

问 AI

Bookmark

View Full Paper