To address challenges such as high susceptibility to environmental noise and ambiguous target characteristics in underwater sonar images, a recognition method integrating multi-dimensional feature fusion with Gaussian process-optimized least squares support vector machines (GPO-LSSVM) is proposed.

First, considering the noise characteristics of sonar images, a target enhancement and segmentation algorithm is developed to highlight regions of interest. Subsequently, multi-dimensional features are extracted, and analysis of variance (ANOVA) and correlation analysis are employed for feature selection and dimensionality reduction. Finally, GPO is introduced to optimize the hyperparameters of the LSSVM, thereby constructing a high-performance classifier.

Experimental results demonstrate that the optimized LSSVM model achieves an accuracy of 80.91% in underwater target recognition, representing a 12.66% improvement over the non-optimized model. While achieving comparable accuracy, the proposed method reduces optimization time by 95.62% and 95.10% compared with particle swarm optimization and deep neural network （DNN） methods, respectively.

The proposed multi-dimensional feature fusion-based target recognition method demonstrates significant effectiveness in enhancing underwater sonar image target identification performance and exhibits considerable practical application potential.