School of Information and Navigation, Air Force Engineering University, Xi'an 710077, China.
School of Information Engineering, Xijing University, Xi'an 710123, China.
Comput Intell Neurosci. 2021 Jan 18;2021:6235319. doi: 10.1155/2021/6235319. eCollection 2021.
Underwater sonar objective detection plays an important role in the field of ocean exploration. In order to solve the problem of sonar objective detection under the complex environment, a sonar objective detection method is proposed based on dilated separable densely connected convolutional neural networks (DS-CNNs) and quantum-behaved particle swarm optimization (QPSO) algorithm. Firstly, the dilated separable convolution kernel is proposed to extend the local receptive field and enhance the feature extraction ability of the convolution layers. Secondly, based on the linear interpolation algorithm, a multisampling pooling (MS-pooling) operation is proposed to reduce the feature information loss and restore image resolution. At last, with contraction-expansion factor and difference variance in the traditional particle swarm optimization algorithm introduced, the QPSO algorithm is employed to optimize the weight parameters of the network model. The proposed method is validated on the sonar image dataset and is compared with other existing methods. Using DS-CNNs to detect different kinds of sonar objectives, the experiments shows that the detection accuracy of DS-CNNs reaches 96.98% and DS-CNNs have better detection effect and stronger robustness.
水下声纳目标检测在海洋探测领域起着重要作用。为了解决复杂环境下声纳目标检测的问题，提出了一种基于扩张可分离密集连接卷积神经网络（DS-CNNs）和量子行为粒子群优化（QPSO）算法的声纳目标检测方法。首先，提出了扩张可分离卷积核，以扩展局部感受野，增强卷积层的特征提取能力。其次，基于线性插值算法，提出了多采样池化（MS-pooling）操作，以减少特征信息丢失并恢复图像分辨率。最后，引入传统粒子群优化算法中的收缩扩张因子和差分方差，利用 QPSO 算法优化网络模型的权重参数。在所提出的方法在声纳图像数据集上进行验证，并与其他现有方法进行了比较。使用 DS-CNNs 检测不同种类的声纳目标，实验表明 DS-CNNs 的检测精度达到 96.98%，并且具有更好的检测效果和更强的鲁棒性。
