基于小波聚类的终端区进场轨迹模式识别  

PATTERN RECOGNITION OF APPROACH LANDING TRAJECTORIES IN TERMINAL AIRSPACE BASED ON WAVELET CLUSTERING

作　　者：王超[1] 郑旭芳 卜宁[1] 

机构地区：[1]中国民航大学空中交通管理学院,天津300300

出　　处：《计算机应用与软件》2016年第11期112-116,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(61039001)

摘　　要：为改善终端区航空器轨迹聚类方法中存在的自动化程度低、无法精确识别异常轨迹的不足,提出基于小波聚类的进场轨迹模式识别方法。首先,建立基于3D空间网格的轨迹相似性矩阵,推导得到轨迹间相似特征子空间,进一步构建轨迹相似特征2D图模型。通过特征图模型的数字化、小波变换与聚类,实现对盛行交通流模式以及异常交通流轨迹的识别。实例分析在无人工指导情况下,从352条进场轨迹中识别出4个类的331条盛行交通流轨迹,以及21条异常轨迹。实验结果证明,该算法克服了目前航空器轨迹聚类领域需要人工确定类数以及难以识别异常轨迹的不足。In order to ameliorate the deficiencies existed in trajectories clustering methods of aircrafts in terminal airspace such as low automation degree and cannot precisely identify the abnormal trajectories,we proposed a wavelet clustering-based pattern recognition method of approach landing trajectories. First we set up the 3D spatial grid-based similarity matrix of trajectories,and obtained through derivation the similar characteristic subspace between trajectories. Furthermore we built the 2D graph model of trajectories similar characteristic. The identification on prevailing traffic flows pattern and abnormal traffic trajectories is implemented through the digitisation,wavelet transformation and clustering on the characteristic graph model. We made analyses on examples,under the condition of no artificial guidance,we identified331 prevailing traffic flows trajectories of 4 categories and 21 abnormal trajectories from 352 approach landing trajectories. Experimental result proved that this algorithm overcomes the deficiency of current aircraft trajectories clustering field that it requires artificial confirmation on the number of categories and is difficult to identify abnormal trajectories.

关 键 词：模式识别 小波聚类 航空器轨迹聚类 异常轨迹识别 

分 类 号：TP301[自动化与计算机技术—计算机系统结构] V355[自动化与计算机技术—计算机科学与技术]