基于资源1号02D高光谱图像湿地水体分类方法对比—

基于资源1号02D高光谱图像湿地水体分类方法对比——以白洋淀为例

陈民, 彭栓, 王涛, 吴雪芳, 刘润璞, 陈玉烁, 方艳茹, 阳平坚

A comparative study of water body classification of wetlands based on hyperspectral images from the ZY1-02D satellite: A case study of the Baiyangdian wetland

CHEN Min, PENG Shuan, WANG Tao, WU Xuefang, LIU Runpu, CHEN Yushuo, FANG Yanru, YANG Pingjian

表5 去除低信噪比波段后基于单点光谱和基于邻域范围光谱的湿地分类精度

Tab.5 Wetland classification accuracy based on pixel-based method and patch-based method after removing low signal-to-noise ratio bands

类型	模型	每类样本的分类准确率					OA/%	mAP/%	F₁	R/%	OA-AB/%
类型	模型	湖泊水面/%	内陆滩涂/%	村庄/%	沼泽/%	水浇地/%	OA/%	mAP/%	F₁	R/%	OA-AB/%
基于单点	SVM	96.68	53.85	96.93	40.63	92.38	80.30	84.00	0.79	86.14	83.90
	AdaBoost	97.86	64.40	97.24	59.22	92.71	86.44	92.80	0.85	90.76	89.48
	RF	97.44	60.11	96.12	53.25	91.44	84.05	90.92	0.83	88.72	85.33
	CNN1D	97.32	85.21	99.45	84.85	95.87	94.30	98.13	0.94	95.94	94.54
	RNN	96.47	81.47	99.45	86.61	96.01	93.12	97.20	0.93	94.55	94.03
	SSTN(Pixel)	94.84	83.40	99.16	85.62	94.56	92.55	95.98	0.92	92.48	93.79
	ViT(Pixel)	95.95	88.38	99.00	79.04	95.53	93.95	97.75	0.93	95.01	95.11
基于邻域	ResNet	96.97	87.69	98.07	92.37	97.54	95.10	98.74	0.95	96.36	96.67
	CNN2D1D	97.61	82.49	99.49	84.62	96.18	93.79	97.99	0.94	96.28	96.03
	CNN3D	95.88	85.54	98.67	92.54	94.62	93.75	97.94	0.94	95.44	96.20
	ViT	95.53	81.71	98.50	85.77	93.96	92.35	96.66	0.92	93.47	98.11
	SSRN	99.88	97.36	99.98	94.21	99.19	99.09	99.95	0.99	99.62	99.19
	SSTN	98.05	95.43	99.65	94.79	97.82	97.51	99.57	0.98	98.02	99.05