Lightweight Algorithm for Detecting Fishing Boats in Offshore Aquaculture Areas Based on YOLOv7-Tiny

• Junhan Peng,
• Xuhong Huang,
• Ronghao Kang,
• Zhihong Chen,
• Jianjun Huang,

Vol. 19, No. 3, March 31, 2025
10.3837/tiis.2025.03.006, Download Paper (Free):

• Deep Learning
• aquaculture farms
• Ship inspection
• YOLOv7-tiny
• lightweight model

Abstract

Offshore aquaculture farms play a crucial role in the production of seafood. However, security concerns, such as theft and illegal fishing activities, remain a significant challenge for these farms. However, the object detection algorithms employed systems for maritime surveillance exhibits significant limitations. For instance, the presence of objects such as bamboo rafts and buoys frequently results in false positives and missed detections. Moreover, the detection accuracy is often compromised when the model is made lightweight. Consequently, there is a necessity to enhance the efficiency and accuracy of detection algorithms while concomitantly reducing its size, a novel network model, PES-YOLOv7, has been developed, building upon and enhancing the YOLOv7-tiny model. The major modifications include replacing the standard 3x3 convolutional kernel with PConv in the backbone network to optimize spatial feature extraction, and incorporating an Exponential Moving Average (EMA) attention mechanism to enhance the model's accuracy and stability. The incorporation of a Slim-neck module has been demonstrated to reduce computational complexity without compromising accuracy. The Focal-EIoU loss function has been introduced to address the challenges posed by complex scenes. The experimental results demonstrate the effectiveness of the enhanced algorithm, achieving a model accuracy of 97.1%, a model size reduction of 30%, and a parameter reduction of 32.2%. The proposed algorithm has been shown to improve the precision of vessel detection in offshore aquaculture farms while achieving a lightweight model. This research can be applied to surveillance systems to better prevent the economic losses caused by the theft of aquatic products.

Statistics

Cite this article