Research on Image Segmentation Algorithm based on Improved SGDF-SSA Algorithm of Chaotic Map and Gaussian Cloud

Abstract

Leaf image segmentation is of great significance in automatically segmenting foreground leaves from noisy background, and the accuracy of image segmentation requires high. Through the improvement of the algorithm, the accuracy of leaf extraction and crop analysis can be greatly improved. The Sparrow Search Algorithm (SSA) is a commonly used image segmentation method. However, the traditional SSA suffers from issues such as a tendency to fall into local optima and insufficient search capability during the optimization process. To address these shortcomings, we propose an improved SSA model, the SGDF-SSA algorithm. First, we use the chaotic phenomenon generated by the SPM chaotic map to initialize the particle population, enhancing the randomness and traversal ability of particles and thereby improving global search capability. Second, we introduce an adaptive Gaussian cloud mutation strategy in the discoverer position update process to further enhance global search ability during iterations. Additionally, we design a sinecosine optimization and inertia weight-based discoverer update mechanism to improve the discoverers’ global search capability. Moreover, we propose a follower update mechanism based on Cauchy chaotic mutation, which combines chaotic mapping and Cauchy mutation to prevent the algorithm from falling into local optima. The improved algorithm outperforms various comparative algorithms in terms of average performance on the CEC2017 benchmark test set, achieving superior results in 11 test functions. It demonstrates better performance in different fidelity parameters and computational time, showing promising potential for plant leaf image segmentation. This advancement is expected to contribute to the progress of leaf pathology analysis and precision agriculture mechanized batch processing technology.