Sustainable Marine Structures

Integrated Application Research on Marine Image Recognition Models

ChihChen Kao

School of Intelligent Manufacturing, Shanghai Zhongqiao Vocational and Technical University, Jinshan, Shanghai, 201514, China

Yufen Peng

Physical Education Office, Yuanpei University of Medical Technology, Hsinchu, Taiwan

Bowen Wu

Professor

DOI: https://doi.org/10.36956/sms.v7i2.1915

Received: 28 March 2025; Published: 6 May 2025

---

Abstract

Marine environments present significant challenges for image processing due to factors such as low light intensity, suspended particles, and varying degrees of water turbidity. These conditions severely degrade the clarity and quality of captured marine images, making accurate image recognition difficult. The problem is further compounded by the limited availability of high-quality, labeled training samples, which restricts the effectiveness of conventional recognition algorithms. Existing techniques in both academic and industrial settings—such as Principal Component Analysis (PCA), Neural Networks, and Wavelet Transforms—typically involve converting color images to grayscale prior to feature extraction. While this simplifies processing, it also results in the loss of essential color information, which is often critical for distinguishing features in marine imagery. To address these issues, this paper proposes a novel approach that preserves and utilizes the full color information of marine images during processing and recognition. The method combines color image representation with Hu's invariant moments to extract stable and rotation-invariant features. These features are then input into a Back Propagation Neural Network (BPNN), which is trained to recognize and classify various marine targets. The integration of color-based feature extraction with BPNN significantly improves recognition performance, particularly under complex environmental conditions. Experimental results show that the proposed system achieves a recognition accuracy exceeding 98%, demonstrating its effectiveness and potential for practical applications in marine exploration, environmental monitoring, and underwater robotics.

Keywords: Marine Image Color Preprocessing; Pattern Recognition; BPNN; Invariant Moments

---

References