基于混合像元分解的高光谱影像柑橘识别方法

发布时间：2018-06-12 14:08

本文选题：柑橘识别 + EO-　；参考：《遥感技术与应用》2017年04期

【摘要】：为及时准确地监测柑橘种植信息,以江西省会昌县作为研究区,采用EO-1 Hyperion高光谱影像作为数据源,构建了基于混合像元分解的高光谱影像柑橘识别方法。首先,针对EO-1 Hyperion高光谱影像提供了242个波段,光谱范围广的特点,在波段选择、大气校正等预处理的基础上,提取研究区典型地物端元光谱曲线;然后,利用全约束线性光谱混合模型进行混合像元分解,提取出柑橘端元的丰度值,并通过对照高分遥感影像,构建柑橘端元丰度与柑橘实际种植的对应的关系。结果表明:由于典型地物端元提取中不可避免的误差及柑橘冠层覆盖度的差异,柑橘种植的准确识别与其柑橘端元丰度阈值存在对应关系。在经过反复试验的条件下,研究区柑橘端元丰度阈值设定在0.30~0.45范围之内,总精度达到90%以上,能够满足柑橘种植识别要求。
[Abstract]:In order to monitor citrus planting information accurately and timely, a hyperspectral image recognition method based on mixed pixel decomposition was constructed in Huichang County, Jiangxi Province, using EO-1 Hyperion hyperspectral image as data source. First of all, aiming at the characteristics of 242 bands and wide spectrum range of EO-1 Hyperion hyperspectral image, based on the pretreatment of band selection and atmospheric correction, the end component spectral curves of typical ground objects in the study area are extracted. The full constrained linear spectral mixing model is used to decompose the mixed pixel to extract the abundance value of the end component of citrus and to construct the relationship between the abundance of citrus end point and the actual planting of citrus by comparing the high score remote sensing images. The results showed that due to the inevitable error in the extraction of the end components of typical ground objects and the difference of canopy coverage, the accurate identification of citrus planting had a corresponding relationship with the threshold of citrus end-element abundance. Under the condition of repeated experiments, the threshold value of end element abundance of citrus in the study area was set in the range of 0.30 ~ 0.45, and the total precision was more than 90%, which could meet the requirement of citrus planting recognition.
【作者单位】：江西理工大学建筑与测绘工程学院;中国科学院遥感与数字地球研究所数字地球重点实验室;
【基金】：国家自然科学基金项目(41561091) 江西省教育厅科技课题(GJJ150659) 江西省社会科学规划课题(14YJ20)资助
【分类号】：S666;TP751

【相似文献】