中文摘要：

无人机视频车辆检测的准确性和稳定性对提取交通信息至关重要。研究专门应用于无人机视频的车辆检测算法，通过对比分析现有车辆检测算法实验结果，结合对称差分能够有效分离背景与目标像素的特点，引入分块建模的思想，提出了基于对称差分分块建模的背景差分法。针对无人机视频车辆检测算法评价指标缺失的现状，分析车辆检测结果，提出正检率、重检率、漏检率、错检率4个评价指标来衡量无人机视频车辆检测算法性能。实验结果表明，所提算法具有较高正检率（均值92．29％，中值100％）和稳定性（四分位距为8％）。上述算法能够在无人机视频中准确地检测车辆。

英文摘要：

The accuracy and stability of vehicle detection based on UAV video is very important for the extraction of traffic information. The vehicle detection algorithms for the UAV video are studied. The detection results of the ex- isting vehicle detection algorithms are compared and analyzed. At the same time, a background subtraction method based on the symmetric difference and block modeling is proposed, by combining the block modeling idea and the symmetric difference which could effectively separate the background and target pixels. In view of the present situa- tion of the lack of evaluation index of vehicle detection algorithm based on UAV video, vehicle detection results are analyzed,4 evaluation indexes, including positive detection rate, repeated detection rate, missing detection rate and false detection rate, are proposed to measure the performance of vehicle detection algorithm based on UAV video. The experimental results show that the background subtraction method based on the symmetric difference and block modeling has high positive detection rate （mean 92. 29% ,median 100% ） and stability （interquartile range 8% ）. The al- gorithm can accurately detect vehicles in the UAV video.