可解释性
人工智能
计算机科学
多光谱图像
遥感
机器学习
模式识别(心理学)
特征(语言学)
特征提取
图像处理
空间分析
上下文图像分类
多光谱模式识别
遥感应用
持续性
网格
空间生态学
高光谱成像
机器视觉
计算机视觉
数据挖掘
深度学习
计算
支持向量机
图像(数学)
共同空间格局
地球观测
空间滤波器
作者
Kexin Liu,Jingyao Wang
出处
期刊:
日期:2026-05-13
卷期号:: 93-93
摘要
This study presents an optical and remote sensing image processing and pattern recognition framework, which applies machine learning algorithms to extract spatial sustainability patterns from multi-source urban imagery, including multispectral and high-resolution optical data. The framework integrates optical imaging techniques with machine learning models, enhancing the accuracy and interpretability of sustainability predictions. Using multi-source remote sensing imagery and spatial datasets, a structured grid-based model is constructed over Stockholm. Seventeen spatial features are extracted via image segmentation, spectral index computation (e.g., NDVI), and object-based feature recognition. An XGBoost regression model is trained to predict sustainability scores for each grid unit, achieving high accuracy (R² = 0.869). To enhance interpretability, SHAP is employed to reveal spatial patterns and quantify the contribution of each image-derived and environmental feature. The predicted results are visualized through heatmaps, highlighting urban sustainability disparities and latent spatial structures. This framework integrates image analysis, interpretable modeling, and spatial pattern discovery to provide data-driven insights for intelligent urban decision-making.
科研通智能强力驱动
Strongly Powered by AbleSci AI