Forest fire detection integrating remote sensing and spatial relationship features
A forest fire classification method integrating multi-source features is proposed after analyzing effects of short-wave infrared bands,terrain correction,and spatial relationship features on burnt area recognition.Terrain correction and short-wave infrared band features are used to reduce effects of mountain shadows and smoke on burnt area recognition,and spatial relationship features are introduced to correct burning area recognitions.The forest fire in Chongqing Municipality in 2022 is studied from Sentinel 2 images and support vector machine model.Application of short-wave infrared bands could reduce interference from smoke,with the recognition accuracy improved by 3％-8％.Terrain correction can reduce interference from mountain shadows,with little effect on recognition accuracy.Introduction of spatial relationship features can significantly reduce interferences from bare ground,hard paved ground,with the accuracy being improved by 3％-4％.The physical mechanism of this method is clear and can effectively extract forest fires(especially burning forest fires).
万方数据
维普
