Scholarlink

<p>The survey of diseases is the prerequisite for the study of the deterioration mechanism of geoheritage and the implementation of targeted protection measures. However, the photogrammetric method currently used mainly for disease investigation heavily relies on personnel experience and has a low level of automation. Therefore, an intelligent model for the typical disease identification of geoheritage based on spectral and image fusion technology is proposed. Firstly, the random forest is used to select the top 3 bands of importance scores to construct a pseudo-color image for feature extraction. Secondly, the extracted image features are fused with spectral features as input vectors to train the model. The findings indicate that integrating both feature types can significantly enhance the model’s accuracy. Subsequently, the identification effects of four models on typical diseases are compared. Finally, the CatBoost model, which performs the best, is used for disease identification of the Niche of Sakyamuni Entering Nirvana. The model proposed in this study, which targets pixel-level classification, can identify typical diseases in geoheritage. Furthermore, the proportion of different disease can be determined by statistically analyzing the ratio of different types of pixels.</p>

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

A Non-Contact Disease Identification Method for Geoheritage Based on the Spectra and Image Fusion Technology

  • 期刊论文

  • 发表日期：

    2025年6月28日

查看全文
  • Xingyue Li,
  • Haiqing Yang,
  • Chiwei Chen,
  • Yongyi Wang,
  • Gang Zhao,
  • Jianghua Ni

摘要

The survey of diseases is the prerequisite for the study of the deterioration mechanism of geoheritage and the implementation of targeted protection measures. However, the photogrammetric method currently used mainly for disease investigation heavily relies on personnel experience and has a low level of automation. Therefore, an intelligent model for the typical disease identification of geoheritage based on spectral and image fusion technology is proposed. Firstly, the random forest is used to select the top 3 bands of importance scores to construct a pseudo-color image for feature extraction. Secondly, the extracted image features are fused with spectral features as input vectors to train the model. The findings indicate that integrating both feature types can significantly enhance the model’s accuracy. Subsequently, the identification effects of four models on typical diseases are compared. Finally, the CatBoost model, which performs the best, is used for disease identification of the Niche of Sakyamuni Entering Nirvana. The model proposed in this study, which targets pixel-level classification, can identify typical diseases in geoheritage. Furthermore, the proportion of different disease can be determined by statistically analyzing the ratio of different types of pixels.