ADS-B历史飞行轨迹数据清洗方法

为有效解决广播式自动相关监视(ADS-B)历史飞行轨迹数据受地面站分布广度、地形阻挡、电磁干扰等影响而出现的各种字段数据异常情况, 建立了ADS-B数据清洗方法, 并将其分为确定清洗对象、字段去重、异常点清理和时间戳修正这4个步骤; 根据已有样本ADS-B历史数据各字段建立了航迹模型并进行有效性分析, 将时间戳、经度、纬度、气压高度和地速等字段定义为特征字段, 并作为清洗对象; 对ADS-B航迹点序列的时间戳、经度和纬度进行去重, 删除数据重复的相邻航迹点; 为提高清洗效率, 使用有噪声的密度聚类(DBSCAN)方法找出特征字段中的离群点, 并进行异常检测与修正; 为使航迹点状态变化符合质点运动学规律, 使用ADS-B航迹点的经度、纬度、气压高度和地速等字段数据修正时间戳, 并存入已扩展的修正后时间戳字段。研究结果表明: 516个样本航班中有97.58%的异常航迹点被有效识别并清理, 清洗后的航迹点状态更具有渐变性特征; 修正前后的总飞行历时存在10~600 s的差异; 时间戳修正效果主要依赖于地速的准确度, 在实际工程中可根据样本航迹的数据特点有选择地使用时间戳修正值; 建立的ADS-B数据清洗方法可为民用航空工程项目中的飞行轨迹分析、评估与计算等方面提供前期数据处理平台。 

Data cleaning method of ADS-B historical flight trajectories

To effectively solve the various field data anomalies in the automatic dependent surveillance-broadcast(ADS-B) historical flight trajectories affected by the ground station distribution breadth, terrain blocking, electromagnetic interference and so on, an ADS-B data cleaning method was established, and implemented by four steps, such as determining the cleaning object, deleting the duplication of field, cleaning the abnormal point and correcting the time stamp. According to the existing sample ADS-B historical data, the track model was established and the validity was analyzed. The fields such as the time stamp, longitude, latitude, pressure altitude and ground speed were defined as the characteristic fields and cleaning objects. The time stamp, longitude and latitude of ADS-B track point sequence were deduplicated to delete the adjacent track points with repeated data. The outliers of characteristic fields were located through the method based on the density-based spatial clustering of applications with noise(DBSCAN) to improve the cleaning efficiency, detect and correct the abnormality. To make the change of track point state conform to the particle kinematic law, the time stamp was corrected by the field data of longitude, latitude, pressure altitude and ground speed of ADS-B track points, and the extended modified time stamp field was saved. Research result shows that 97.58% of the abnormal track points in the 516 sample flights are effectively identified and cleaned. The cleaned track point state changes more smoothly. The total flight duration before and after correction varies between 10-600 s. The correction effect of time stamp mainly depends on the accuracy of ground speed. The corrected time stamp should be selectively used according to the data characteristics of sample track in practical engineering applications. The established cleaning method of ADS-B data can provide a preliminary data processing platform for the trajectory analysis, evaluation and computing in civil aviation engineering projects.