應用Watershed影像分割法於高解析QuickBird衛星影像進行道路區塊之萃取

Road Extraction Based on Watershed Segmentation for High Resolution QuickBird Satellite Images

張立雨(L. Y. Chang) ； 陳繼藩(C. F. Chen)

《航測及遙測學刊》 11卷3期 (2006/09) Pp. 261-267

https://doi.org/10.6574/JPRS.2006.11(3).3

摘要

近年來，資源衛星影像之空間解析度大幅提升至公尺等級，因此，在此類高解析影像上，可辨識之地面特徵當然更為豐富，也更為複雜。然而，也因此提高了自動化影像判釋與特徵萃取技術之複雜度及難度。在本文中，將提出利用Watershed影像分割之方法來萃取道路區塊之資訊。首先，利用簡化之影像融合運算融合多光譜及全色態影像資訊並應用數學型態學侵蝕運算降低影像之複雜度。第二，針對前一步驟所產生之紅外線波段影像，以Watershed影像分割之方法將影像進行分割並配合適當之區塊合併程序以得到適當之區塊化影像。第三，計算分割後之影像區塊上之各波段光譜平均值，並同時以ISODATA影像分類法，由光譜資訊分類出合適之道路區塊。最後，配合數學型態學閉合運算及面積門檻值進行道路區塊二次過濾。研究成果顯示，本文所提出之方法可由QuickBird高解析衛星影像萃取出適用之道路區塊，其精確度約為70%。

關鍵字

道路區塊萃取 ； 影像分割 ； 高解析衛星影像

並列摘要

Recently, the spatial resolution of earth observation satellites is significantly increased to a few meters. However, it is more difficult to develop automated image algorithms for automated image feature extraction and pattern recognition. In this study, we propose a scheme to extract road information from high resolution satellite images based on watershed segmentation. First, the multi-spectral and panchromatic images are fused by simplified image fusion technique and the mathematical morphology erosion is applied next to decrease the complexity of fused image. Second, the segmentation result is created by applying watershed segmentation on the infrared image generated in previous step. The iteratively patches merging procedure is also performed to prevent from over segmentation. Third, we take multi-spectral information for each segment by averaging the pixel values of corresponding segment area of multi-spectral image. The ISODATA classification algorithm is followed to classify the suitable road patches from the spectral information of each segment. Finally, the mathematical morphology close operation and area threshold are used as a final refinement.
In this study, we use panchromatic and multi-spectral images of QuickBird satellite as test dataset. The experiment result shows that the accuracy of generated road objects by proposed scheme is about 70%.

並列關鍵字

Road Patch Extraction ； Image Segmentation ； High Resolution Satellite Images