长时序高光谱图像清晰度影响因素分析

目的 清晰度是评价对地观测成像仪影像数据质量的典型指标之一,可以反映成像仪对地物边缘变化的敏锐程度。已有的对地观测成像仪在轨测试及图像质量评价方法研究中,往往关注遥感影像清晰度是否达标,或监测其变化趋势,未对清晰度变化影响因素进行深入探讨。针对这一问题,本文主要对长时间序列的成像仪成像清晰度的变化以及影响因素进行探讨。方法 以天宫一号高光谱成像仪短波红外谱段0级数据作为研究对象,首先利用改进的基于边缘检测的清晰度算法计算出影像的清晰度,其次将各影像数据对应的成像仪工程参数进行筛选,然后利用Apriori算法对长时间序列高光谱影像的清晰度与成像时刻的工程参数进行关联规则挖掘,利用最小支持度阈值和最小置信度阈值筛选出强关联规则,并附加提升度和余弦对强关联规则进行验证,最后结合3维散点图对影响清晰度的主要因素进行定量分析。结果 经大量测试数据表明,天宫一号高光谱成像仪短波红外谱段影像清晰度较好,影响清晰度的主要因素有太阳高度角、拍摄积分时间以及平台稳定性(包括俯仰角、偏航角和滚动角的稳定性)。太阳高度角与图像清晰度呈正相关关系,即当太阳高度角大于65°时,影像清晰度较高,当太阳高度角小于30°时,影像清晰度较低;平台稳定性与图像清晰度呈正相关关系,即当太阳高度角大于30°且小于65°时,平台稳定性高倾向于得到清晰度较高的图像,平台稳定性低倾向于得到清晰度较低的图像;拍摄积分时间与图像清晰度呈负相关关系。结论 基于关联规则挖掘的长时序高光谱图像清晰度影响因素分析方法是一种有效的分析方法,可以挖掘出与影像清晰度强关联的工程参数。后续可扩大工程参数范围,利用此分析方法进一步研究遥感图像其他指标与工程参数的关联关系。

Analysis of factors influencing clarity of hyper spectral images in long time series

Objective The clarity of remote sensing images is one of typical evaluation indicators for assessing the image quality of Earth-observation payload. The sharpness degree of imaging payload can be determined from the changing gray values at the edge of objects. Currently however, studies on orbit tests of Earth-observation payload and quality evaluation of remote sensing images focus mostly on whether the clarity values of the remote sensing images reach a certain standard. Very few of these studies conduct further analysis on the factors influencing the clarity of a remote sensing image. To address this issue, we assessed the clarity values of short-wave infrared (SWI) images of Tiangong-1 hyper-spectral payload in a long time series and analyzed the influencing factors on clarity. Method First, we calculated clarity using an improved clarity algorithm, which is based on image-edge detection. Then, the corresponding Tiangong-1 hyperspectral payload engineering parameters of the remote sensing images were obtained. Finally, association rule mining was performed using the typical Apriori algorithm on the clarity values of the SWI images of Tiangong-1 hyperspectral payload in the long time series and their corresponding engineering parameters. The association rules were viewed as alternative strong association rules when their supporting and confidence measurements exceeded the set minimum thresholds. Only the alternative strong association rules verified by the addition of the lift interest and cosine interest measurements could be adopted. The influencing factors on the clarity of the SWI images of Tiangong-1 hyper spectral payload were analyzed using the strong association rules and three-dimensional scatter plots. Result After testing a considerable amount of data, the results show that the clarity values of the SWI images of Tiangong-1 hyperspectral payload are good. The influencing factors on the clarity of the SWI images of Tiangong-1 hyperspectral images include solar angle, integral time, and platform stability (including the stability of pitch, yaw, and rolling angles). Solar angle is positively correlated with clarity. When the solar angle is larger than 65°, the clarity values tend to be higher, but when the solar angle is less than 30°, the clarity values tend to be lower. When the solar angle is larger than 30° and less than 65°, platform stability is positively correlated with clarity. Integral time has a negative relationship with remote sensing image clarity. Conclusion The method of analyzing the factors influencing the clarity of SWI images of Tiangong-1 hyperspectral payload in a long time series is feasible and effective. The relationships between the clarity values of the SWI images of Tiangong-1 hyper-spectral payload and their corresponding engineering parameters have been identified. The scope of engineering parameters should be expanded to identify quantitative and more significant relationships. Furthermore, the association rule mining method could be helpful in analyzing the influencing factors for other remote sensing image evaluation indicators.