基于数据筛选的无人机测绘数据异常检测

为降低无人机测绘数据异常检测的误检率、漏检率，缩短检测时间，文章提出一种基于数据筛选的无人机测绘数据异常检测方法。采用支持向量机对无人机测绘数据进行数据流分块、竖向规范化处理及时间切片处理等预处理；基于卷积神经网络分析数据确定数据潜在规律；采用无监督聚类算法对数据进行聚类，利用滑动窗口处理得到数据流簇心因子并进行聚类；根据判断标准对异常数据分块处理，确定是否存在异常因子；采用重叠累加值计算方法对异常数据点进行筛选，完成无人机测绘数据异常检测。实验结果表明:与传统卷积神经网络方法相比，采用该方法对异常数据检测，其误检率降低了约11%、漏检率降低约8.1%，并且检测时间缩短了11.3 min。

Anomaly Detection Method of UAV Surveying and Mapping Data Based on Data Filtering

Aiming at the problems of high false detection rate, high missed detection rate and long detection time of traditional methods, an anomaly detection method of UAV surveying and mapping data based on data screening is proposed. Support vector machine is used to preprocess the UAV mapping data, such as data flow blocking, vertical normalization processing and time slicing processing. Data is analyzed based on convolution neural network to determine the potential law of data. The unsupervised clustering algorithm is used to cluster the data, and the sliding window is used to get the cluster center factor of the data flow and cluster it. Block the abnormal data is rejected according to the judgment criteria to determine whether there are abnormal factors. The overlapping cumulative value calculation method is used to screen the abnormal data points to complete the anomaly detection of UAV mapping data. The experimental results show that compared with the traditional convolution neural network method, the false detection rate of abnormal data detection by this method is reduced by about 11%, and the missed detection rate is reduced by about 8.1%, and the detection time is shortened by 11.3 min.