Enhanced walking palm tree optimization for high-quality multilevel thresholding of RGB histogram-based color satellite image segmentation

Multilevel thresholding for color satellite image segmentation is challenging due to spectral variability, noise, atmospheric distortions, randomness, multiple regions of interest, and ambiguous boundaries. Although various nature-inspired algorithms have been developed for color image segmentation, they often yield suboptimal results when applied to complex satellite images, especially as the number of threshold levels increases. To address these challenges, this paper proposes an Enhanced Walking Palm Tree Optimization (EWPTO) algorithm, combined with minimum cross-entropy, for high-quality multilevel thresholding of RGB color satellite images. To improve the performance of the original Walking Palm Tree Optimization (WPTO) algorithm, the proposed method integrates the strategies of enhanced initialization, adaptive exploration, geometric update rules, diversity preservation, and guided opposition learning, which offer better convergence speed and segmentation accuracy. The proposed approach is applied for forest fire detection & monitoring applications, and it is evaluated using eight RGB satellite forest fire images across multiple threshold levels, and the performance is compared against eleven state-of-the-art optimization algorithms, including GWO, WOA, HHO, SMA, GJO, HBO, BES, CSA, WHO, MGO, and MSO. The experimental results demonstrate that the EWPTO algorithm exhibits superiority in achieving a PSNR of 32.76 dB, SSIM of 0.995, an FSIM of 0.999, NCC of 0.994, and reducing MSE by 17% compared to the best competitor. Convergence analysis demonstrates that the proposed EWPTO algorithm achieved faster convergence and stable optimization behaviour. The CEC-2022 benchmark test suite was used to estimate the global optimization capability of the proposed EWPTO method, and the Wilcoxon signed-rank test and Friedman mean-rank statistical analysis ensure superiority and robustness of the proposed EWPTO method. Overall the proposed method EWPTO is suitable for image segmentation applications, including land cover classification, vegetation analysis, urban planning, environmental monitoring, and water body detection.