基于AIS数据的交汇水域船舶会遇态势辨识
Ship Encounter Situation Recognition by Processing AIS Data from Traffic Intersection Waters
-
摘要: 船舶会遇态势辨识是实现船舶安全避碰的重要前提,交汇水域复杂的通航环境和船舶运动的不确定性使在辨识会遇态势过程中容易出现误判的情况。为准确快速辨识交汇水域的船舶会遇态势,提出联合支持向量机-贝叶斯滤波(Support Vector Machine and Bayesian Filter,SVM-BF)的辨识会遇态势模型和算法。基于会遇几何模型计算两船的相对距离和航向差作为会遇特征,并将特征序列输入模型,在SVM初辨识的基础上,使用BF器对SVM产生的初判结果进行累积判别,提高辨识不同会遇态势的准确性。选取长江口南槽交汇水域的船舶自动识别系统(Automatic Identification System,AIS)数据,进行会遇过程分析与特征提取和SVM-BF模型训练和算法验证。结果表明:基于SVM-BF模型辨识会遇态势算法能针对交叉、对遇和追越等态势进行及时有效识别,与仅采用SVM的辨识模型相比,SVM-BF能有效降低误判概率。Abstract: The SVM-BF(Support Vector Machine and Bayesian Filter) are combined to construct the model and algorithm for ship encounter situation recognition. The relative distances and bearings between ships are calculated according to the geometry of ship distribution and defined as the encounter feature set. The feature sequences are fed to the support vector machine to do preliminary classification. The output of the SVM is further processed by the Bayesian filter to make the final results more reliable. The AIS(Automatic Identification System) data from the south channel of the Yangtze estuary is adopted to verify the model and algorithm. The experiments indicate that the recognition model and algorithm can handle typical encounter situations, such as crossing, head-on and overtaking.