An edge detection technique has been proposed to facilitate the automatic alignment of boarding bridges with aircraft doors, thereby addressing issues associated with manual operation and potential collisions during the conventional docking process. To enhance the robustness of noise suppression, pixel value difference-based weights have been introduced as weighting factors in the median filtering process, wherein pixels assigned higher weights exert a greater influence on the median calculation. Threshold segmentation has been performed using Otsu’s method based on maximum inter-class variance, enabling the automatic determination of an optimal threshold and thereby improving the accuracy of aircraft door feature extraction. Furthermore, the Sobel operator has been modified by integrating a weighted average of neighboring pixel values, which significantly enhances its edge detection performance. Experimental results have demonstrated that the application of weighted median filtering improves the peak signal-to-noise ratio (PSNR) of the image to 37.93 dB, representing a 13.95% increase compared to conventional median filtering. The modified Sobel operator has been shown to increase the number of detected edge points by 134.5% relative to other operators, with improvements of 25.2% in precision, 193.6% in recall, and 136.1% in the F₁ score. These results provide meaningful insights into the optimization of edge detection methodologies for aircraft door localization.