节点文献
基于高光谱图像特征提取与凸面几何体投影变换的目标探测
OBJECT DETECTION BASED ON FEATURE EXTRACTION FROM HYPERSPECTRAL IMAGERY AND CONVEX CONE PROJECTION TRANSFORM
【摘要】 在介绍凸面几何体概念和高光谱图像光谱线性模型的基础上,探讨了凸面几何体和高光谱图像的关系.提出了在光谱特性提取的基础上利用凸面几何体投影变换进行高光谱图像目标探测的方法,并成功地应用于亚运村建材市场屋顶板材和亚运村中心地区真假草坪的自动识别和探测中.试验证明该方法不需要探测目标的任何先验知识就能达到比较好的目标探测效果.同时试验还表明,由于高光谱图像波谱范围宽,特别是有短波红外、中红外和热红外范围的图像,不仅增加了地物的可分性,而且为探测地物物理性质提供了可能.
【Abstract】 After introducing the conception of convex geometry and spectral linear model, the relationship between convex cone and hyperspcetral imagery was discussed. A method based on feature extraction and convex cone projection transform for hyperspectral imagery object detection was proposed. The method was tested with hyperspectral data obtained from OMIS Ⅱ. Based on the experiments. the capability of detecting object’s physical property by hyperspectral imagery was also discussed.
【Key words】 hyperspectra; target detection; feature selection; projection transformation;
- 【文献出处】 红外与毫米波学报 ,Journal Infrared Millimeter and Waves , 编辑部邮箱 ,2004年06期
- 【分类号】P237
- 【被引频次】51
- 【下载频次】1008