高光谱遥感目标探测主要利用目标和背景的光谱特征差异进行目标识别。一般情况下,影像的空间和光谱分辨率越高,探测效果越好。但多数情况下空间和光谱分辨率难以同时满足需求。针对该问题,本文利用Field Imaging Spectrometer System(FISS)地面高光谱成像仪器,通过在稀疏草地上布设人工绿色目标,研究了目标和背景光谱相似情况下,单一均匀背景下小目标探测问题,提出空间和光谱尺度定量分析方法,得到目标探测适用的空间和光谱尺度。结果表明:(1)利用FISS高光谱仪器进行人工目标探测,所需的空间分辨率约为目标尺寸的2倍以内;(2)当光谱分辨率优于40 nm时,目标和背景的两个主要特征:反射峰的位置和波段趋势差异均可被描述,在原始空间分辨率5倍(0.85 cm)以内,探测精度可以达到0.94以上。由于反射峰间距20 nm,当光谱分辨率低于40 nm时,该特征消失,造成探测精度的下降;(3)当光谱分辨率低于40 nm时,选取目标、背景光谱特征差异较大的波段可提高探测的有效性,在舍弃目标背景相似波段后,探测精度上升,得到本实验的最佳波段组合为红、绿、蓝、黄及红边波段。
Hyperspectral target detection is based on the spectral characteristic difference between the target and the background. Generally, the finer the resolution is, the higher the accuracy will be. Yet the spatial and spectral resolution can barely meet the need simultaneously, for technical obstacles. A tradeoff between the two factors is needed for effective target detection and the choice of appropriate remote sensing data for target detection has always been concentrated nowadays. Most previous studies focused on the precision assessment of the target detection in a particular case using current image spectrometer data, but failed to provide a quantified criterion for either the spatial or the spectral resolution appropriate for the data in that case. This study focused on the quantification of the scale impact of spectral and spatial resolution on target detection precision, with respect to situations where the target is small and has a similar spectrum to the homogenously distributed background.Here we proposed a technical method for spectral and spatial resolution assessment for target detection, setting green context and sparse grass as the target and background, respectively. Through the down sampling processing of the high spatial hyperspectral image from Field Imaging Spectrometer System(FISS) together with Constrained Energy Minimization(CEM)detection algorithm and Receiver Operation Characteristic(ROC) evaluation method, this study analyzed the relationship between the spatial and spectral resolution and the detection accuracy. And then it proposed the optimal spatial and spectral scale for target detection. Results revealed that:(1) With the decline of spatial resolution, the detection accuracy experienced three stages of descending rates:gently-dramatically-gently. The corresponding spatial resolution before the second stage is the effective scale for detection. Using FISS data(4-7 nm spectral resolution and 1.4 nm sampling interval), the required spatial resolution for target detection was about within twice the size of the target;(2) When the spectral resolution was finer than 40 nm, two main features:the reflection peaks and basic trend differences, associated with the target and the background, could be identified. The detection accuracy would reach 0.94 or above within the spatial resolution of 0.85 cm. When the spectral resolution was coarser than 40 nm, the differences of reflection peaks disappeared since they were 20 nm apart and the detection accuracy decreased;(3) Given spectral resolution insufficiency(>40 nm), the basic R,G,B bands added the yellow and red edge bands appeared the optimal combination for target detection with respect to current multispectral remote sensors. It was concluded that the quantitative analysis method and results of spatial and spectral scales for target detection would be of great significance for both data source selection and studies on other target-background combinations under similar conditions.