基于视频数据的机场跑道外来物检测

提出了一种基于视频数据的机场跑道外来物(FOD,Foreign Object Debris)检测算法,该算法包括几何校正、背景差分、杂波抑制和伪装消除等4个步骤.其中,几何校正排除了摄像头轻微抖动造成的图像差异;背景差分利用每个像素对应的颜色向量信息,建立背景模型并进行定时更新;杂波抑制和伪装消除利用差分图像和原始图像信息,建立马尔科夫随机场和概率统计模型,在降低虚警的同时提高了检测率.将本算法应用于两组不同光照条件下获取的视频图像,检测不同形状和颜色的FOD目标;采用本算法对机场实测数据进行测试,验证了算法的有效性.

Foreign object debris detection for airport runway with video data

Foreign object debris (FOD) detection algorithm was proposed for airport runway with video data, including four steps of geometric adjustment, background subtraction, noise suppression and camouflage elimination. Firstly, geometric adjustment rejected the image change due to slight camera motion. In the step of background subtraction, the color vector information to each pixel was used for the establishment of the background model and its periodical update. For noise suppression and camouflage elimination, Markov random field and probability statistical model were established based on the information from the subtracted and original images, which reduced the false alarms and improved the detection rate. The algorithm was applied to two sets of live video images under different lighting conditions, with the FOD targets of different shapes and colors. The algorithm effectiveness was also approved by the tests on the airport ground-truth data.