摘要
提出一种基于偏振光谱二向反射分布函数图像的物质自动分类方法,该方法主要选择偏振光谱二向反射分布函数信息作为新的特征用于物质自动分类.采用支撑向量机的分类方法对不同的天气条件(晴天、多云、阴天)下处于杂乱的自然草地背景环境中的典型目标进行分类,最后比较三种不同特征选择对于分类准确度的影响.采取三种不同的特征选取方法,分别为采用单一的光谱特征、偏振光谱特征及偏振光谱二向反射分布函数特征.最后通过实验得出:将偏振光谱二向反射分布函数作为分类特征在三种不同的天气情况下,分类准确度都较高,特别是在阴天天气条件下,分类准确度明显高于其它两种特征选择.即使是在阴天低照度下的场景中,当不同目标和背景之间的灰度很接近时,采用本文方法也能准确的进行自动分类.
A new classify method based on spectropolarimetric BRDF imagery is proposed. The performances of three different selected features in classifyication results under various weather conditions including sunny sky, cloudy, and dark sky are emphasized. The three selected features are material spectral information, spectropolarimetric information, and spectropolarimetric BRDF information respectively. Support Vector Machine method is used to classify targets in clutter grass environments, then the classify results based on spectropolarimetric BRDF features three different weather conditions respectively. are compared with the other two features under the The results show that the method based on spectropolarimetric BRDF features performs the best among the three, no matter what the weather conditions are, and its advantage shows most evidently especially in the dark sky. Selecting the spectropolarimetric BRDF information as features in the materials classification will enhance the precision at most time,even in the case when the gray values between backgrounds and targets are very near.
出处
《光子学报》
EI
CAS
CSCD
北大核心
2010年第6期1026-1033,共8页
Acta Photonica Sinica
基金
国家自然科学基金(60602056)
国家自然科学基金重点资助项目(60634030)
高等学校博士学科点专项科研基金(20060699032)
航空科学基金(2007ZC53037)
西北工业大学英才计划基金资助
