Aerial images are photographs or images captured from an elevated vantage point above the Earth surface. They are typically taken from aircraft, drones, satellites, or other aerial platforms. Aerial image segmentation is a fundamental task in computer vision and remote sensing. They are necessary for different applications, such as urban planning, land use and land cover classification, surveying and mapping, environmental monitoring, agriculture, disaster management, military, and others. Aerial image segmentation helps extract valuable information from images and can lead to more accurate analysis and interpretation. This paper presents a multilevel thresholding aerial image segmentation technique based on a modified Osprey Optimization Algorithm (OOA) based on a multi-strategy mechanism. In the modified OOA (MOOA), we applied the Double attractors to enhance the exploration of OOA. The dynamic random search mechanism is applied to enhance its exploitation. Therefore, the multi-strategy mechanism improves the performance of OOA and supports it to avoid premature convergence. Sixteen aerial images are used to validate the performance of the MOOA. It is also compared to well-known multilevel thresholding methods. The results show the high ability of the MOOA to allocate the best threshold values that improve the quality of segmented images according to different performance metrics. For example, PSNR, FSIM, and SSIM are 30.0503, 0.98489, and 0.93648, respectively.
Elaziz et al. (2026) studied this question.