基于改进MobilenetV2网络的声光图像融合水下目标分类方法*

针对小样本条件下水下目标分类准确率低、计算资源量大的问题，提出一种声光图像融合目标分类方法。首先，对MobilenetV2网络进行改进，去掉第9层之后的网络层，并将该层卷积通道数改为128，通过Flatten层进行数据降维，增加一个全连接层得到分类结果；其次，设计一种融合网络结构，将声光图像成对输入网络进行特征提取，在中间层利用通道拼接算法实现特征图融合，使用融合特征进行目标分类。在真实数据集上对网络进行训练，结果表明，改进的MobilenetV2网络对水下目标的分类性能更好，融合网络的分类准确率相比融合前有所提高，更加适用于水下目标分类任务。

Acoustic-optical Image Fusion Underwater Target Classification Method Based on Improved MobilenetV2

To solve the problems of low accuracy and high consumption of underwater target classification under the condition of small samples, an acoustic-optical image fusion classification method is proposed. Firstly, the mobilenetv2 network is improved to reduce the network complexity and improve the network fitting effect. Secondly, a fusion network structure is designed, which inputs acoustic image and optical image in parallel for feature extraction. Then, the extracted feature maps are combined in the middle layer and the real data are used to train the network. Results show that the average accuracy of the improved mobilenetv2 network is 1.1% higher than before in the classification experiments, the parameters and calculation are reduced by 44% and 23% respectively. Using different algorithms at different locations, the classification accuracy respectively improved 4.9%, 4.4%, 5.2%, 5.9%, 5.8% and 6.0% compared with pre fusion network, indicating that the performance of the fusion network in classification accuracy and stability are improved further.