基于深度学习的多旋翼无人机单目视觉目标定位追踪方法

针对无人机对目标的识别定位与跟踪，本文提出了一种基于深度学习的多旋翼无人机单目视觉目标识别跟踪方法，解决了传统的基于双目摄像机成本过高以及在复杂环境下识别准确率较低的问题。该方法基于深度学习卷积神经网络的目标检测算法，使用该算法对目标进行模型训练，将训练好的模型加载到搭载ROS的机载电脑。机载电脑外接单目摄像机，单目摄像头检测目标后，自动检测出目标在图像中的位置，通过采用一种基于坐标求差的优化算法进行目标位置准确获取，然后将目标位置信息转化为控制无人机飞行的期望速度和高度发送给飞控板，飞控板接收到机载电脑发送的跟踪指令，实现对目标物体的跟踪。试验结果验证了该方法可以很好的进行目标识别并实现目标追踪

Monocular vision target tracking method for multi-rotor Unmanned Aerial Vehicle based on deep learning

Aiming at the target recognition, location and tracking of UAV(Unmanned Aerial Vehicle), this paper proposes a multi-rotor uav monocular vision target recognition and tracking method based on deep learning, which solves the problems of high cost of traditional binocular camera and low recognition accuracy in complex environment.This method is based on the target detection algorithm of deep learning convolutional neural network, which is used to conduct target model training and load the trained model into the onboard computer equipped with ROS.Onboard computer external monocular camera, monocular camera detecting target, the automatically detect the target in the image position, by adopting a kind of optimization algorithm based on coordinate for poor get target location accurate, then the target position information into control of unmanned aerial vehicle (uav) flight speed and height expectation for flight control board, flight control board accepting follows commands sent to the airborne computer, realize the target object tracking.Experimental results show that this method can recognize and track targets well