A Skyline Extraction Algorithm Based on Vertical Local Contrast Measurement

The skyline serves as a natural reference landmark, which can be utilized for confirming aircraft attitude and scene-matching navigation. To meet the stringent requirements for speed and robustness in airborne applications, this letter proposes a novel skyline extraction method based on the dynamic programming (DP) algorithm. First, a grayscale image is obtained from the raw RGB image through color space transformation and subsequent Gaussian filtering. Then, an LLGM map is generated from the grayscale image using the locally longitudinal gradient measurement (LLGM) designed in this study. Subsequently, the LLGM map undergoes sequential filtering, hierarchical thresholding, and morphological processing, resulting in a binary map containing refined skyline candidate pixels. Finally, the DP algorithm is introduced to search for the position of the skyline curve in the binary map. Through experimental tests in different complex scenarios, the robust extraction ability of the proposed skyline extraction method is verified. Furthermore, the final qualitative analysis and statistical results demonstrate that the proposed method achieves fast and accurate extraction compared to state-of-the-art methods. In particular, the intermediate test results reveal the reasons behind the improved speed and accuracy of the DP search.