2D Hyperspectral Satellite Image-based Detection of Forest Fires

该建议将研究和开发从卫星图像数据自动探测森林火灾的新方法。将采用机器学习方法，训练探测系统根据专家提供的森林火灾标签和卫星获取的高光谱二维图像数据识别森林火灾的图像模式特征。将使用标准检测准确性指标（例如检测特异性和灵敏度以及曲线下面积（AUC）测量）对系统进行评价。

This proposal will research and develop novel methods for automatic detection of forest fires from satelliteimage data. A machine learning approach will be adopted, whereby a detection system will be trained torecognize image patterns characteristic of forest fires, based on forest fire labels provided by an expert humanand hyperspectral 2D image data acquired by satellite. The system will be evaluated using standard detectionaccuracy metrics, e.g. detection specificity and sensitivity and the area-under-the-curve (AUC) measure.