基于局部光谱空间分析的多尺度边缘检测

分析了传统边缘检测算法在多光谱影像应用中的局限性，提出了一种基于局部区域的光谱空间平均半径测度的边缘检测方法，该方法扩展了原有的单波段影像边缘检测的概念，将像元的灰度信息转化为光谱矢量，提高了检测的可靠性，并给出了多尺度边缘组合的算法。通过MAIS成像光谱仪对鄱阳湖30个波段数据试验，证明了此方法对边缘检测的有效性。

MULTI-SCALE EDGE DETECTION BASED ON SPECTRAL SPACE ANALYSIS OF LOCAL AREA

This paper analyzes the limitation of traditional edge detection methods in the use of multispectral images, and proposes an edge detection technique based on spectral space analysis of local area whose average radius can be calculated. This method expands the previous concept of edge detection of the single image and, because of the transformation from the gray value to the gray vector and the consideration of multi-scale combination, can acquire satisfying results.