IARC第7代任务中定位与目标跟踪方法研究
发布时间：2018-10-09 21:26
【摘要】：空中机器人在航拍、运输、巡检、搜救等领域有广泛应用,其定位、导航和目标跟踪技术也成为研究热点。国际空中机器人大赛是最具挑战性的无人机赛事之一,大赛目的是推动空中机器人技术发展,其任务都是当前无人机技术所不能达到的,参与和研究比赛任务有重要意义,本文在大赛第7代任务背景下展开,主要工作和创新如下所述。1.根据比赛任务,搭建了智能四旋翼平台。平台利用机载电脑实现视觉定位、目标跟踪、避障等上层功能,针对遥控和程序控制的切换问题,设计了手自动切换系统,系统由STM32单片机接收遥控器和机载电脑指令,根据遥控器手自动切换档位,将相应遥控或程序控制指令传给下层飞控。该四旋翼平台既能应用于比赛场景,也可用于巡检、救援等其他场合。2.基于单目视觉和IMU开发了定位与悬停系统。该系统利用光流法对飞行器运动速度进行估计,并结合卡尔曼滤波算法实现了光流速度与IMU信息的数据融合,提高了飞行器速度估计的精度和平滑性,同时基于速度信息设计了位置速度串级控制器,实现了飞行器的视觉悬停。此外,针对光流法积分存在累计误差的问题,提出了一种基于比赛场地方格和张正友相机标定法的视觉定位方案,详细阐述了方案原理,并进行了可行性分析。3.针对单目标识别、跟踪与触碰问题,提出了一种综合解决方案。该方案利用CamShift算法进行单目标识别,利用PID控制器进行目标跟踪,并采用基于开关控制的高度控制器实现触碰。在AR.Drone四旋翼上进行了实验,验证了算法的有效性,实现了比赛内容中的单目标驱赶目标。4.针对比赛场地存在多个运动目标的情况,设计了基于移动目标颜色特征及最近邻算法的多目标检测器与跟踪器,提出了基于移动目标颜色区域形状特征的目标朝向检测算法。实验表明了算法的有效性。相关实验视频在优酷频道(http://i.youku.com/u/UMzMwMDg3NTAlMg==/videos)中,可下载浏览。
[Abstract]:Aerial robots are widely used in aerial photography, transportation, inspection, search and rescue, and so on. The international aerial robot competition is one of the most challenging UAV competitions. The purpose of the competition is to promote the development of aerial robot technology, and its tasks are beyond the reach of the current UAV technology. It is of great significance to participate in and study the competition tasks. This paper is carried out in the context of the 7th generation task, the main work and innovation described below. 1. According to the task of the competition, the intelligent four-rotor platform is built. The platform uses the airborne computer to realize the upper functions such as visual positioning, target tracking, obstacle avoidance and so on. Aiming at the switching problem of remote control and program control, a hand-automatic switching system is designed. The system receives remote control and airborne computer instructions by STM32 microcontroller. According to the remote control hand automatically switch gear, the corresponding remote control or program control instructions to the lower flight control. The four-rotor platform can be used not only in competition scenes, but also in inspection, rescue and other occasions. 2. A positioning and hovering system is developed based on monocular vision and IMU. The system uses the optical flow method to estimate the velocity of the vehicle, and combines the Kalman filter algorithm to realize the data fusion of the optical flow velocity and the IMU information, which improves the accuracy and smoothness of the vehicle velocity estimation. At the same time, the position speed cascade controller is designed based on the velocity information, which realizes the visual hovering of the aircraft. In addition, aiming at the problem of accumulative error in the integration of optical flow method, this paper presents a visual positioning scheme based on the grid of competition site and the calibration method of Zhang Zhengyou camera. The principle of the scheme is described in detail, and the feasibility analysis is made. Aiming at the problem of single target recognition, tracking and touching, a comprehensive solution is proposed. In this scheme, CamShift algorithm is used for single target recognition, PID controller is used for target tracking, and height controller based on switch control is used to realize contact. Experiments are carried out on the AR.Drone four-rotor, the validity of the algorithm is verified, and the single-target drive target .4in the content of the competition is realized. In view of the situation that there are many moving targets in the competition field, a multi-target detector and tracker based on moving object color feature and nearest neighbor algorithm is designed, and a target orientation detection algorithm based on the shape feature of moving target color region is proposed. Experiments show the effectiveness of the algorithm. The video about the experiment is available for download on Youku Channel (http://i.youku.com/u/UMzMwMDg3NTAlMg==/videos).
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41
[Abstract]:Aerial robots are widely used in aerial photography, transportation, inspection, search and rescue, and so on. The international aerial robot competition is one of the most challenging UAV competitions. The purpose of the competition is to promote the development of aerial robot technology, and its tasks are beyond the reach of the current UAV technology. It is of great significance to participate in and study the competition tasks. This paper is carried out in the context of the 7th generation task, the main work and innovation described below. 1. According to the task of the competition, the intelligent four-rotor platform is built. The platform uses the airborne computer to realize the upper functions such as visual positioning, target tracking, obstacle avoidance and so on. Aiming at the switching problem of remote control and program control, a hand-automatic switching system is designed. The system receives remote control and airborne computer instructions by STM32 microcontroller. According to the remote control hand automatically switch gear, the corresponding remote control or program control instructions to the lower flight control. The four-rotor platform can be used not only in competition scenes, but also in inspection, rescue and other occasions. 2. A positioning and hovering system is developed based on monocular vision and IMU. The system uses the optical flow method to estimate the velocity of the vehicle, and combines the Kalman filter algorithm to realize the data fusion of the optical flow velocity and the IMU information, which improves the accuracy and smoothness of the vehicle velocity estimation. At the same time, the position speed cascade controller is designed based on the velocity information, which realizes the visual hovering of the aircraft. In addition, aiming at the problem of accumulative error in the integration of optical flow method, this paper presents a visual positioning scheme based on the grid of competition site and the calibration method of Zhang Zhengyou camera. The principle of the scheme is described in detail, and the feasibility analysis is made. Aiming at the problem of single target recognition, tracking and touching, a comprehensive solution is proposed. In this scheme, CamShift algorithm is used for single target recognition, PID controller is used for target tracking, and height controller based on switch control is used to realize contact. Experiments are carried out on the AR.Drone four-rotor, the validity of the algorithm is verified, and the single-target drive target .4in the content of the competition is realized. In view of the situation that there are many moving targets in the competition field, a multi-target detector and tracker based on moving object color feature and nearest neighbor algorithm is designed, and a target orientation detection algorithm based on the shape feature of moving target color region is proposed. Experiments show the effectiveness of the algorithm. The video about the experiment is available for download on Youku Channel (http://i.youku.com/u/UMzMwMDg3NTAlMg==/videos).
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41
【参考文献】
相关期刊论文 前3条
1 曹美会;鲜斌;张旭;文曦;;基于视觉的四旋翼无人机自主定位与控制系统[J];信息与控制;2015年02期
2 张梁;徐锦法;;基于双目视觉的无人飞行器目标跟踪与定位[J];计算机工程与应用;2014年24期
3 倪磊;曾庆化;庄f，
本文编号：2260849
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2260849.html
