城市更新背景下融合深度学习的非正式 绿地数字识别技术研究进展

城市更新过程中，非正式绿地在为城市提供环境、 社会和生态效益方面发挥着至关重要的补充作用。然而，非正 式绿地由于数量繁多、面积较小、种类多样、边界模糊等特征 导致其难以高效识别。深度学习在分割图像进行识别及自主学 习方面有极大的优势。因此，基于深度学习，对多源数据依托 下的非正式绿地识别研究进行综述。首先,基于非正式绿地内 涵归纳其识别特征与识别难点；其次，梳理识别非正式绿地常 用的4种数据类型，即基于深度学习识别的高分辨遥感影像、 街景图片，用于辅助识别的网络媒体数据、PPGIS平台数据； 再次，对现有深度学习进行非正式绿地整体布局识别、耦合多 源数据进行辅助识别的方法研究进展进行归纳总结；最后，对 未来非正式绿地空间识别应用与发展提出展望与建议

Advances in Informal Green Space Digital Recognition Technology Based on Deep Learning in the Context of Urban Renewal

In the process of urban renewal, informal green spaces play a crucial complementary role in providing environmental, social, and ecological benefits for cities. However, informal green spaces are difficult to efficiently identify due to their large number, small area, diverse types, and blurry boundaries. Deep learning has great advantages in segmenting images for recognition and autonomous learning. Therefore, a review is being conducted on the use of multi-source data to identify informal green spaces based on deep learning. Firstly, it summarizes recognition features and difficulties based on the connotation of informal green spaces. Secondly, it sorts out the four commonly used data types for recognition of informal green spaces, namely high-resolution remote sensing images, street view images based on deep learning recognition; online media data, PPGIS data for auxiliary recognition. Thirdly, it summarizes the research progress of existing deep learning methods for identifying informal green spaces and coupling multi-source data for auxiliary identification methods. Finally, prospects and suggestions are proposed for the future application and development of informal green space recognition.