Scholarlink

This case presents an AI-driven method for identifying pocket park installation (PPI) sites in high-density areas. Using street view images (SVIs) and convolutional neural networks (CNNs), it develops four parallel-task deep learning models (assessing Space, Vitality, Facility, Pleasurability) trained via pairwise comparisons and Trueskill algorithms. Applied in Beijing within the Fourth Ring Road, the approach processes 122,309 sampling points to classify site potential. Results show higher PPI potential in the old city than expanded zones. A “Potential-Priority” framework then integrates supply-demand analysis to prioritize sites, overcoming traditional 2D limitations and enabling fine-grained, 3D-informed urban park planning.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Pilot CaseII—AI-Based Identification of Pocket Parks’ Installation Sites

  • 章节

  • 发表日期：

    2025年

查看全文
  • Conghui Zhou

摘要

This case presents an AI-driven method for identifying pocket park installation (PPI) sites in high-density areas. Using street view images (SVIs) and convolutional neural networks (CNNs), it develops four parallel-task deep learning models (assessing Space, Vitality, Facility, Pleasurability) trained via pairwise comparisons and Trueskill algorithms. Applied in Beijing within the Fourth Ring Road, the approach processes 122,309 sampling points to classify site potential. Results show higher PPI potential in the old city than expanded zones. A “Potential-Priority” framework then integrates supply-demand analysis to prioritize sites, overcoming traditional 2D limitations and enabling fine-grained, 3D-informed urban park planning.