无人机在高层建筑灭火中的视觉算法研究

Research on the visual algorithm of UAV in high-rise buildings firefighting

在高层建筑灭火救援中,针对传统飞行控制器在GPS受高温干扰时远程目标瞄准易失效等缺点,提出Kalman滤波算法与均值漂移相结合的算法实现对火焰目标快速准确识别,利用Kalman滤波器进行辅助位置预测,可以在一定程度上解决目标尺寸变化过大或被遮挡等情况导致的目标丢失或误判问题。实验结果表明,本文算法可以一定程度上降低高层无人机灭火视觉识别失效率,其位置预测的稳定性、精度、抗干扰能力基本都满足要求,对于扑救初期火灾具有一定意义。

In high-rise building firefighting, traditional UAV has disadvantages such as when the GPS is under the interruption of high temperature, the remote target aiming is prone to failure.Against the problems, algorithm combing Kalman filtering algorithm and mean shift was proposed to identify the flame fast and accurately. Predicting the location by Kalman wave filter auxiliary can solve the problems of target missing or mistaking caused by oversize target or blocked target to a certain degree. The experiment results showed that, the algorithm can reduce the misjudgment of visual recognition to some extent,the location prediction can satisfy the requirement of stability, accuracy and anti-jamming capability, which is of great significance to firefighting in early stage.