基于GF1卫星的红树林病虫害遥感识别

定量化红树林病虫害监测评估可为红树林保护修复提供科学依据。以广西北仑河口红树林保护区的红树林为研究对象,利用高分一号(GF1)卫星数据建立模型识别受虫害红树林,结果表明:GF1数据的红光和近红外波段是判识红树林健康与否的优选波段;差值植被指数(difference vegetation index,DVI)是区分健康和虫害红树林的优选植被指数,通过训练样本可划定不同群落虫害红树林DVI阈值区间;利用归一化植被指数(normalized difference vegetation index,NDVI)和绿光波段可有效去除水体和裸露滩涂等背景干扰信息,综合DVI、NDVI和绿光波段,构建虫害红树林识别决策树模型,模型识别精度良好。木榄-白骨壤、秋茄-白骨壤、木榄-桐花树、木榄-白骨壤+桐花树群落的综合指标F(F-measure)大于0.89,且准确率A(accuracy)、精确率P(precision)、召回率R(recall)大于0.85,但白骨壤分类精度较低,综合指标仅0.76。

Mangrove pests recognition based on GF1 satellite data

Quantitative monitoring and evaluation of mangrove pests can provide a scientific basis for mangrove protection and restoration. In this paper, mangroves in the Beilun River Mangrove Reserve of Guangxi Province were taken as the research object, and a model was established to identify the mangroves pests using GF1 satellite data. The results showed that: (1) The red and near-infrared bands of GF1 satellite data are the preferred bands to judge whether the mangroves are healthy or not. (2) The Difference Vegetation Index (DVI) is the preferred vegetation index to distinguish the healthy mangroves from the pest-infested mangroves, the DVI threshold of pest-infested mangrove for different communities can be determined by training samples. (3) Water body and bare beach can be effectively removed by NDVI and green band with proper thresholds. Then, a decision tree model for pest-infested mangrove recognition was established using DVI, NDVI and green band. The recognition accuracy is satisfactory. The comprehensive index F of four communities including Bruguiera gymnoihiza-Avicennia marina, Kandelia obovate-Avicennia marina, Bruguiera gymnoihiza-Aegiceras corniculatum, Bruguiera gymnoihiza-Avicennia marina+Avicennia marina were higher than 0.89, moreover, and the accuracy rate A, accuracy rate P, and recall rate R were higher than 0.85. However, the classification accuracy of Avicennia marina was low with F was only 0.76.