论文部分内容阅读
为研究约束条件下目标探测的尺度效应,采用重采样的方法对高光谱图像进行处理,得到了空间分辨率不断降低而光谱分辨率不断增加的系列图像,而后用JGM距离对目标和背景的可分性进行分析,并用RXD异常检测算法验证探测率和虚警率.结果表明:当光谱分辨率较高,而空间分辨率很低时,混合像元造成的不确定性会导致目标无法探测;当光谱分辨率较低,而空间分辨率较高时,虽然目标光谱分辨率很低,但是目标和背景像元不存在融合,目标不存在亚像元级情况,目标的探测率会很高;在光谱分辨率和空间分辨率都无法达到最高的情况下,最佳探测结果对应的图像空间分辨率和光谱分辨率是两者的折中,既不是光谱分辨率最高的图像,也不是空间分辨率最高的图像.
In order to study the scale effect of target detection under constrained conditions, a hyperspectral image is processed by resampling method, and a series of images with decreasing spatial resolution and increasing spectral resolution are obtained. Then, the distance between target and background using JGM The results show that when the spectral resolution is high and the spatial resolution is low, the uncertainty caused by the mixed pixels will lead to the target can not be detected; When the spectral resolution is low and the spatial resolution is high, although the target spectral resolution is low, there is no fusion between the target and the background pixels, the target does not exist in the sub-pixel level, and the target detection rate will be very high. In the case that the spectral resolution and the spatial resolution can not reach the maximum, the spatial resolution and the spectral resolution corresponding to the best detection result are the compromises of both, and are neither the images with the highest spectral resolution nor the spatial resolution The highest rate image.