报告详情
Satellite-Based Oil Spill Classification Using a Hybrid Stacking Ensemble
编号：192
访问权限：仅限参会人
更新：2026-11-10 17:06:21 浏览：13次
Online
报告开始：2026年11月20日 14:30（Asia/Kolkata）
报告时间：15min
所在会场：[S4] Artificial Intelligence and Intelligent Computing [S4-2] Artificial Intelligence and Intelligent Computing
暂无文件
摘要
Ocean oil spills can be considered as one of the
worst environmental impacts on the coastal ecosystem, fishery
and health of people. Nevertheless, although there are many
possibilities to observe wide ocean zones using remote sensing
technology, a huge amount of data collected every day by
satellite cannot be processed manually. Nowadays, deep learning
techniques prove their good performance, but they require
expensive GPU hardware and time-consuming training processes.
Besides, deep learning uses the black box approach. In order
to deal with these disadvantages, the lightweight classification
pipeline based on handcrafted features and CPU training was
developed. Satellite images were used as an input to the classifier
and contained 87 handcrafted features which were divided into
four categories: textures calculated with the help of GLCM ap
proach, statistics, spectra and frequency domain features. Using
Optuna framework, optimal hyperparameters were found for
five machine learning algorithms: SVM, RF, XGBoost, CatBoost
and HistGradientBoosting (HistGBM). Probabilities estimated by
these models were aggregated in the MLP meta-classifier for final
classification task. Furthermore, SHAP values for each model
were calculated in order to assess the influence of every feature
on the prediction result. Accuracy and AUC scores of the pipeline
reached 98.67% and 99.81% correspondingly on the dataset of
10,650 satellite images.
关键词
Satellite Imagery
报告人
Joyabi J
Student Karunya Institute of Technology and Sciences稿件作者
全部评论
重要日期
-
会议日期
11月20日
2026
至11月22日
2026
-
10月31日 2026
初稿截稿日期
-
11月07日 2026
提前注册日期
-
11月22日 2026
注册截止日期
主办单位
United Societies of Science
承办单位
Pondicherry University
协办单位
IEEE Section
IEEE Madras Section
IEEE Madras Section
联系方式
- ICon INDIA 2026
- ic******@usssociety.org
发表评论