多色彩高空间分辨率遥感影像矢量边缘信息提取算法与应用

Vector Edge Extraction from Multi-Spectral High Spatial Resolution Remotely Sensed Imagery(MHSRRSI) Based on Multi-Color space

相对单波段灰度影像而言,多波段高空间分辨率遥感影像中可用于边缘检测的光谱信息更加丰富。鉴于Canny算子在灰度图像边缘检测中的优越性能,本文利用输出融合策略对其适用于高空间分辨率遥感影像矢量边缘检测作了改进。基于可视化开发平台VC++.NET,编程实现了福州市航拍的高空间分辨率遥感影像红绿蓝三个标准波段在RGB、IHS、Y IQ、YUV、C IELUV色彩空间中对各种地物矢量边缘信息的有效提取。对高空间分辨率遥感影像矢量边缘各分量的分析认为,由于波谱范围差异的影响,在上述色彩空间中不同地物类型边缘检测时响应程度具有显著的不同。本文研究结果表明,该算法参数设置和色彩空间选择对高空间分辨率遥感影像矢量边缘信息提取有较大的影响。

MHSRRSI has much more rich spectral information for detecting edges than single band imagery,and thus flexibly selecting tactics on edge detection and extraction according to feature properties is more important.Since its advantages in detecting edges from monochrome imagery,Canny operator is improved by means of the strategy of output fusion so as to be suitable for detecting vector edges from MHSRRSI.The improved Canny algorithm contains the following steps.Firstly,in order to effectively wipe off geometrical details and mini-spot noise in the inner of geographical semantic feature object,we apply the Gaussian smooth filter to the original imagery.Secondly,we get the multi-dimension gradient vector imagery based on the smoothed imagery.Thirdly,we utilize the method of non-maximum restraining to acquire the primitive vector edge imagery based on the gradient vector imagery.At last,we use the technique of couple thresholds to gain the finally vector edges from the primitive vector edge imagery.Based on Visual VC＋＋.NET platform,the algorithm is implemented to effectively extract vector edges in the multi-color space from the study area in Fuzhou.Because of discrepancies among the spectral range of those components of MHSRRSI,it will result in remarkable differences of response extent of edge detection among different kinds of geographical features.The response extent of edge detection among different kinds of geographical features in multi-color space is also analyzed.The conclusions of this article benefit feature selection for edge detection among different kinds of geographical features and the subsequent edge extraction.Meanwhile,it indicates that the parameter and the selection of color space will give remarkable impacts on the extraction of vector edges.