Article: Urban green space landscape planning and design based on superpixel segmentation algorithm Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.40 pp.100 - 121 Abstract: With the acceleration of urbanisation, urban green space planning faces challenges such as poor domain adaptability and low calculation efficiency. To solve these problems, this paper proposes a deep learning model based on superpixel segmentation algorithm. The model achieves high-precision green space segmentation by fusing domain-independent embedding, lightweight structure and multi-modal fusion mechanism. Moreover, the model adopts an end-to-end framework, which sequentially performs data preprocessing, domain-independent depth super-pixel segmentation, feature enhancement and multi-task loss optimisation. The experimental results show that the model achieves 95.2% accuracy and 89.7% mIoU on multiple global datasets. At the same time, the performance degradation rate of the model under domain offset is only 5.3%, the number of parameters is less than 0.5 M, and the inference time of edge devices is 45.2 ms, which is significantly better than the traditional algorithm. Therefore, this study innovatively combines cross-domain robustness with lightweight design to provide a scalable solution for urban green space planning. In the future, multi-sensor fusion and adaptive training will be used to improve practicality. Inderscience Publishers - linking academia, business and industry through research

Title: Urban green space landscape planning and design based on superpixel segmentation algorithm

Authors: Dingzhen He

Addresses: School of Design and Art, Liuzhou Institute of Technology, China

Abstract: With the acceleration of urbanisation, urban green space planning faces challenges such as poor domain adaptability and low calculation efficiency. To solve these problems, this paper proposes a deep learning model based on superpixel segmentation algorithm. The model achieves high-precision green space segmentation by fusing domain-independent embedding, lightweight structure and multi-modal fusion mechanism. Moreover, the model adopts an end-to-end framework, which sequentially performs data preprocessing, domain-independent depth super-pixel segmentation, feature enhancement and multi-task loss optimisation. The experimental results show that the model achieves 95.2% accuracy and 89.7% mIoU on multiple global datasets. At the same time, the performance degradation rate of the model under domain offset is only 5.3%, the number of parameters is less than 0.5 M, and the inference time of edge devices is 45.2 ms, which is significantly better than the traditional algorithm. Therefore, this study innovatively combines cross-domain robustness with lightweight design to provide a scalable solution for urban green space planning. In the future, multi-sensor fusion and adaptive training will be used to improve practicality.

Keywords: superpixel segmentation; urban green space; landscape planning; design.

DOI: 10.1504/IJICT.2026.153314

International Journal of Information and Communication Technology, 2026 Vol.27 No.40, pp.100 - 121

Received: 09 Dec 2025
Accepted: 30 Jan 2026
Published online: 01 May 2026 *

Title: Urban green space landscape planning and design based on superpixel segmentation algorithm

Keep up-to-date