【摘要】 海洋声学目标探测技术在海洋资源开发、海洋生物保护、区域海洋生态监测及军事领域等方面都具有无可替代的优势．然而，接收传感器常常会受到外界干扰声源的影响，从而导致定位性能下降．为解决这一问题，本文提出了一种基于优化随机森林的海基站水下声学定位模型来改善定位准确度．首先，对采集的原始信号进行数据预处理．对于信号无干扰采集数据，仅保留有效的样本，剔除包括缺失、异常、相同或相似的数据．对于信号干扰下采集数据，先对异常情况进行分类，然后去除存在干扰的基站数据．利用主成分分析对数据进行降维，并将综合特征作为高斯混合聚类的输入．为确定最优超参数，采用网格搜索和交叉验证的封装模块并设置混合系数，然后构建基于优化随机森林的水下声学定位模型，并将其与多点空间定位模型进行对比．实验结果表明，在信号无干扰和有干扰条件下，采用多点空间定位模型计算出目标三维定位误差RMSE分别为5.648 m、26.264 m，而采用本文模型预测的RMSE分别为0.232 m、4.354 m．所提出的方法显著提高了目标的定位精度，并具有较强的泛化能力．且在无干扰条件下，对于x、y及z方向的估计，所提出的方法的RMSE分别为0.122 m、0.099 m和0.171 m．而在有干扰条件下，对于x、y及z方向的估计，RMSE分别为1.715 m、1.691 m和3.627 m．特别是z方向的定位误差最大，与实际情况相吻合．以上数据结果验证了所提出方法的有效性，为水下运动目标高精度定位及定位导航系统的论证设计提供了新的思路．更多还原

【Abstract】 The ocean acoustic target detection technology offers unique advantages in the development of marine resources,protection of marine life,regional marine ecological monitoring,and military applications. However,the performance of receiving sensors is often compromised by external interference sources,leading to decreased positioning accuracy. To address this issue,this paper proposes an optimized random forest-based underwater acoustic positioning model for marine base stations to improve the positioning accuracy. First,data preprocessing is performed on the collected raw signals. For data acquired without interference,unnecessary samples,including missing,abnormal,identical,or similar data,are removed. For data acquired with interference,abnormal cases are identified,and the data from the base stations causing interference is eliminated. Subsequently,principal component analysis(PCA) is used for dimensionality reduction,and the synthesized features are used as inputs for Gaussian mixture clustering. Furthermore,the optimal hyperparameters are determined through a combination of grid search,cross-validation,and mixture coefficients. Then,an underwater acoustic positioning model is established using optimized random forest,and the estimation effect is compared with the multi-base spatial positioning model. Experimental results demonstrate that the root mean square errors(RMSEs) of the three-dimensional positioning of the multi-base spatial positioning model are 5.648 m and 26.264 m respectively under the condition of no interference and interference,while the RMSEs of the model proposed in this paper are 0.232 m and 4.354 m respectively. The proposed method significantly improves the positioning accuracy of the target and exhibits strong generalization capabilities. Specifically,the estimated RMSEs along the x,y,and z directions with no interference are 0.122 m,0.099 m,and 0.171 m,respectively,while the estimated RMSEs along the x,y,and z directions with interference are 1.715 m,1.691 m,and 3.627 m,respectively. In particular,the largest positioning error occurs in the z direction,which is consistent with the actual situation. These results validate the effectiveness of the proposed method,providing new insights for high-precision positioning of underwater moving targets and the design of positioning navigation systems.更多还原