Multi-scale target detection algorithm in circular synthetic aperture sonar images based on improved YOLOv8 model
摘要
文章提出一种基于改进 YOLOv8模型的圆周合成孔径声呐(circular synthetic aperture sonar,CSAS)图像目标多尺度自适应检测算法.首先,设计多尺度特征增强模块,利用多分支卷积与跨尺度特征融合机制,增强模型对小目标的表征能力;其次,设计注意力机制,结合通道压缩和自注意力查询策略,动态增强目标特征响应;此外,针对实际声呐数据较为稀缺的问题,基于神经网络设计数据增强方法,生成符合声呐成像特性的仿真图像数据,有效扩充样本集,提升训练数据的多样性.实验结果表明,所提检测算法较基准模型的平均检测精度提升了 6.3个百分点,达到89.7%,验证了该算法对CSAS图像目标检测的有效性.
This paper proposes a multi-scale adaptive target detection algorithm for circular synthetic aperture sonar(CSAS)images based on an improved YOLOv8 model.First,a multi-scale feature enhancement module is developed,which leverages multi-branch convolutional operations and cross-scale feature fusion mechanisms to strengthen the model's capability in characterizing small and indistinct targets.Second,an attention mechanism is designed,integrating channel compression and self-attention query strategies to dynamically enhance the feature response of target regions.Furthermore,to mitigate the scarcity of real-world sonar data,a neural network-based data augmen-tation method is introduced,generating synthetic training images that adhere to sonar-specific imaging characteristics,thereby effectively expanding the dataset and enhancing the diversity of the training data.Experimental results demonstrate that the proposed algorithm achieves a mean Average Precision(mAP)of 89.7%,surpassing the baseline model by 6.3 percentage points,and validates its effectiveness in detecting CSAS images.