A Feature Fusion Airport Detection Method Based on the Whole Scene Multispectral Remote Sensing Images

Being one of the most important infrastructures, airports play a vital role in both civil fields and military fields. However, detect airports directly based on the whole scene remote sensing images (RSIs) with complex background remains challenging. To address this issue, this article proposes a method that mainly combines spectral features and geometric features of airports with concrete runways to detect multiple airports simultaneously from a whole scene multispectral image with medium-high spatial resolution and with comparatively few bands (contains blue, green, red, and near-infrared bands). Specifically, a decision tree algorithm was developed based on the analysis of spectral features to extract main concrete areas within the whole RSI. Then, the geometric features are used to aim at extracting the point marks of candidate airports. The influence of different image spatial resolutions of the proposed method is explored and the detection effect and processing efficiency of proposed method is verified based on whole scene RSIs with complex background. The analysis of experimental results shows that Sentinel-2 images is more suitable for airport detection than Gaofen-6 and Landsat-8 images based on the proposed method. In addition, the proposed method provides high-accuracy detection of category Ⅳ airports based on Sentinel-2 images with different background complexity in experimental areas indicate the proposed method has a high robust and a good applicability. Finally, run-time test of the proposed method was conducted, and it demonstrates the proposed method has the higher processing efficiency when applying to regional airport detection.