Abstract:

The Flight Deck Aviation Support Job（FDASJ） on the large ship is the preparatory work happening on the flight deck. The detection of FDASJ benefits to analyze and improve efficiency. During the action of the air fleet, the FDASJs are always performed for all of the aircrafts simultaneously, involving hundreds of operators, high level instruments and facilities. Aiming at the difficulties that to detect multi-FDASJs occurring in the same scene during the same time period, participants of which are located alternatively, it presents a FDASJs detection method based on Multi-entity Bayesian Network（MEBN）. In the first place, it obtains the clear trajectories for all the objects in the video during the preprocessing stage. In the FDASJ detection stage, with the input of the trajectories, it first performs the automatic constrained clustering algorithm to cluster the trajectories participating in the same FDASJ into the same cluster; then infers the type of the FDASJ inside each cluster using the FDASJ recognition algorithm based on MEBN. Experiments on the video data set captured from the flight deck demonstrate the effectiveness of the proposed method.

Key words: Flight Deck Aviation Support Job（FDASJ）, complex event detection, automatic clustering, constrained clustering, Multi-entity Bayesian Networks（MEBN）

摘要：

大型舰船舰面保障作业指驻扎于大型舰船的飞机的一次出动过程中，发生于舰面的针对飞机的出动回收准备活动（如飞机调运、物资转运、飞机滑跑起飞等）。基于舰面监控视频检测舰面保障作业，利于分析并提高保障作业效率。在大型舰船携带的机群出动过程中，各飞机的保障作业同时进行，涉及数以百计的操作员、高级器械及设施设备的协作。针对大型舰船舰面保障作业检测问题中，同一视频场景内同时发生多个保障作业且参与者位置交叉、重叠的难点，提出一种基于多实体贝叶斯网络的大型舰船舰面保障作业检测技术。该技术假设在视频数据预处理阶段获取视频中各目标物清晰连贯的运动轨迹；在保障作业检测阶段，输入目标物轨迹，先基于带约束的轨迹自动聚类算法，自动计算聚类个数，并将参与同一保障作业的轨迹聚至同一类；在各聚类内，采用基于MEBN的保障作业识别技术，推断保障作业类型。大型舰船舰面监控视频数据集上的实验表明，该方法对同一视频场景中同时发生的且参与者位置交叉的舰面保障作业具有良好的检测效果。

关键词: 大型舰船舰面保障作业（FDASJ）, 复杂事件检测, 自动聚类, 带约束的聚类, 多实体贝叶斯网络（MEBN）