Abstract: The popularity of low cost and small UAVs brings a series of serious problems and is difficult to supervise. Moreover, due to the disturbance of environmental objects, camera’s jitter and sampling noise, the existing method can detect small targets such as UAVs in low accuracy under visible light. In order to solve the above problems, a UAVs detect method is proposed by random forest. This method collects the image sequences under visible light, and detects small moving targets in the image by using gaussian?mixture?model and clustering detection algorithm. Then, the targetis discriminated to fuse target multi features by using random forest algorithm, and finally the detection target is obtained. The experimental results show that this method can detect small target of unmanned aerial vehicle and improve the accuracy of detection.

Key words: multi features, random forest algorithm, target detection

摘要： 随着低成本小型无人机的普及带来了一系列的严重问题并且难以监管。并且，由于环境物体的扰动、摄像机的抖动及采样噪声等因素导致现有方法在可见光图像下对无人机等小目标检测准确率低。针对上述问题，提出了一种基于随机森林的无人机检测方法。该方法采集可见光下的图像序列，使用混合高斯模型和聚类检测算法检测图像中的运动小目标，继而通过随机森林算法融合目标的多种特征进行目标的判别，最终得到检测目标。实验结果表明，该方法可有效地检测出无人机运动小目标并大幅提高检测的精确率。

关键词: 多特征联合, 随机森林算法, 目标检测