| 基于计算机视觉的无人机物体识别追踪 |
| |
| 引用本文： | 李 杰,刘子龙.基于计算机视觉的无人机物体识别追踪[J].教育技术导刊,2020,19(1):21-24. |
| |
| 作者姓名： | 李 杰 刘子龙 |
| |
| 作者单位： | 上海理工大学 光电信息与计算机工程学院，上海 200093 |
| |
| 基金项目： | 国家自然科学基金项目（61573246） |
| |
| 摘 要： | 提出一种将ROS系统的AR.Drone飞行器作为载体，基于飞行器实现物体识别追踪的具体优化方案。在计算机视觉方面，结合物体识别OpenCV模块中的Haar级联分类器与卡尔曼滤波，实现无人机的目标识别以及对错误目标的过滤功能，使飞行器在搭载摄像头模块后，可结合现有视觉模型完成目标识别要求，算法融合后的系统性能具有良好的鲁棒性；在飞行器控制方面，结合飞行器自身的反馈控制模块与基于相对位置控制的PD位置控制器，优化飞行器自身姿态及目标追踪过程中的动态参数调节优化功能，使飞行器在目标追踪过程中具有良好的自适应性。基于以上两点优化方案建立实验模型，取得了较好的实验效果。具体相对位置估计均方根误差实验结果为：在x方向上为0.124 5m，在y方向上为0.243 7m，在z方向上为0.176 8m，证明了该优化方案的实用性。
|
| 关 键 词： | 计算机视觉 无人机 目标识别 目标追踪 ROS系统 OpenCV |
| 收稿时间： | 2019-04-08 |
Detection and Tracking for UAV Based on Computer Vision |
| |
| LI Jie,LIU Zi-long.Detection and Tracking for UAV Based on Computer Vision[J].Introduction of Educational Technology,2020,19(1):21-24. |
| |
| Authors: | LI Jie LIU Zi-long |
| |
| Institution: | School of Optical-Electrical and Computer Engineering， University of Shanghai for Science and Technology，Shanghai 200093，China |
| |
| Abstract: | In this paper， an AR.Drone aircraft of ROS system is taken as a carrier， and the specific optimization scheme of object recognition tracking based on aircraft is proposed. In computer vision， combined with the Haar cascade classifier and Kalman filter in the object recognition OpenCV module to realize the UAV recognition target and the function of filtering the wrong target， the aircraft combines the existing visual model after the camera module is equipped. The target recognition requirements are completed， and the system performance after the algorithm fusion has good robustness. In the aspect of aircraft control， combined with the feedback control of the aircraft’s own module and the PD position controller based on relative position control， the aircraft’s own attitude and dynamic parameter adjustment and optimization during target tracking are optimized， and the attitude and optimization function of the aircraft’s own attitude and target tracking process are also optimized so that the aircraft has good adaptability in the target tracking process. Based on the above two optimization schemes， this paper establishes an experimental model and achieves better experimental results. The experimental results of the specific relative position estimation root mean square error are 0.124 5m in the x direction， 0.243 7m in the y direction， and 0.176 8m in the z direction， which proves that the proposed optimization scheme has practicality. |
| |
| Keywords: | computer vision UAV target detection target tracking ROS system OpenCV |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
