Multi-level Segmentation of Chilli Images Driven by Walrus Optimization Algorithm with Two Strategies

High-precision image segmentation is beneficial for meeting the requirements of precision agricultural management. In this regard, an enhanced walrus optimization algorithm (LE-WaOA) is proposed, combined with minimum cross-entropy method for multi-level segmentation of chilli images. LE-WaOA integrates lifespan-based Lévy flight and elite group genetic strategy, enhancing optimization convergence and accuracy. The smaller the cross-entropy, the more refined the segmentation of the chili pepper images. By minimizing the cross-entropy between the segmented and original images, LE-WaOA aims to find the optimal set of threshold combinations for the highest segmentation accuracy. The smaller the cross-entropy, the more detailed segmentation the chilli images present. Comparative experiments on CEC2017 and real chilli images demonstrate the superiority of LE-WaOA over DE, CMAES, and other metaheuristic algorithms. LE-WaOA achieves the lowest cross-entropy and performs excellently in the peak signal-to-noise ratio evaluation metric.