Scholarlink

This paper presents a project for the recognition and classification of forest fires in multi-spectral satellite images, characterizing them, using the CBIR technique and a data augmentation based on the user’s experience. The images are from the State of San Luis Potosi in Mexico, and the 3 challenge is to present a first analysis and characterization with this type of images, to try to make 4 a supervised classification of when a forest fire is present or not. We worked with a limited set of 5 images and thanks to the proposed technique we were able to sufficiently characterize them. In a first 6 experimentation with a distance classifier it was not possible to classify the images with fires, and in 7 a second stage a genetic algorithm was developed and implemented to search for the most relevant 8 bands, achieving a 100% classification. The results shown here match the criteria of environmental specialists.

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

Fire Risk Assessment in San Luis Potosí’s Middle Zone Using CBIR and Evolutionary Computation Techniques for Land Image Analysis

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • José Rodrigo Torres-Licona,
  • Juan Villegas-Cortez,
  • César Augusto Puente-Montejano,
  • Salomón Cordero-Sanchez

摘要

This paper presents a project for the recognition and classification of forest fires in multi-spectral satellite images, characterizing them, using the CBIR technique and a data augmentation based on the user’s experience. The images are from the State of San Luis Potosi in Mexico, and the 3 challenge is to present a first analysis and characterization with this type of images, to try to make 4 a supervised classification of when a forest fire is present or not. We worked with a limited set of 5 images and thanks to the proposed technique we were able to sufficiently characterize them. In a first 6 experimentation with a distance classifier it was not possible to classify the images with fires, and in 7 a second stage a genetic algorithm was developed and implemented to search for the most relevant 8 bands, achieving a 100% classification. The results shown here match the criteria of environmental specialists.