摘要
为了利用大气向下红外辐射进行云的地基红外遥感,分析了晴空辐射阈值的确定方法以及有云天空和晴空的LBP谱,进一步提出了阈值与纹理相结合的云识别方法。辐射阈值法可以逐像元地进行云识别,但一旦阈值确定不准,会导致云识别误差较大;有云与晴空的LBP谱有明显差异,可以利用晴空相似度阈值进行云识别,但由于计算LBP谱需要在较大区域进行,使得纹理法无法实现像元级的云识别。利用纹理法区域有云无云识别结果可以调整晴空辐射阈值,进而提高阈值法云识别准确性。初步分析表明,该方法是可行的。
To retrieve cloud from the atmospheric downward infrared radiance between 8～14 micrometer,a new cloud identification method combining the threshold method with texture classification was proposed.Every pixel could be identified as cloud based on the threshold method,but it was difficult to set a threshold level suitable for all weather conditions due to the varied water vapor and visibility.The LBP spectrum of the image of the cloud sky is clearly different from that of the clear sky.So the LBP spectrum can be used to identify clear sky.But the LBP spectrum can only be used to identify cloud in the whole sky or larger area rather than identify cloud on pixel.The threshold level was adjusted according to the recognition result based on LBP spectrum.A new threshold level was set until the recognition result based on the threshold was in accordance with that based on LBP spectrum.The results show that the method combining threshold with texture effectively increases cloud recognition rate.
出处
《解放军理工大学学报（自然科学版）》
EI
北大核心
2011年第4期397-402,共6页
Journal of PLA University of Science and Technology(Natural Science Edition)
关键词
云识别
阈值
纹理
LBP谱
cloud identification
threshold
texture
LBP spectrum
