基于旋转FHOG-LBP特征旋翼无人机动态检测算法

针对旋翼无人机通过视觉准确、快速检测其他旋翼无人机存在的问题,提出基于旋转FHOG-LBP动态检测算法。首先,针对旋翼无人机特有的外形和运动状态,通过形成的样本库建立外部结构模型；其次,将方向梯度直方图(HOG)进行傅里叶变换(Fourier),使其具有快速旋转不变性,在局部二进制模式(LBP)上加入角度旋转偏移值,进行串行融合得到旋转FHOG-LBP特征,使用支持向量机递归特征消除算法(SVM-RFE)进行训练,并使用滑动窗口的检测算法对目标进行检测；最后,通过无人机动态目标测试集进行了实验,实验结果表明,提出的动态检测算法比传统方法精度和时效提高。因此,该方法可以解决具有动态变化或旋转变化的目标检测困难的问题。

Dynamic Detection Algorithm of Rotor UAV Based on FHOG-LBP Feature

In order to solve the problem that the rotor UAV can accurately detect the existence of other rotor UAV, an rotate FHOG-LBP dynamic detection algorithm is proposed. Firstly, for the unique shape of the rotor UAV, the external structure model is established through the formed sample library. Secondly, Fourier transform of direction gradient histogram (HOG) is improved to have fast rotation invariance in the local binary mode (LBP). The angle rotation offset value is added, and the final rotation invariant FHOG-LBP feature is obtained by serial fusion. The support vector machine recursive feature elimination algorithm (SVM-RFE) is used for training, and the sliding window detection algorithm is used to detect the target. Finally, experiments were carried out through the UAV dynamic target test set. The results show that the method can improve the accuracy and timeliness of rotor UAV over conventional methods. It is concluded that the method can solve the problem of difficulty in target detection with dynamic changes or rotational changes.