基于MEANSHIFT的小型无人机位置解算方法研究

【作者】 王晓红；

【导师】 成怡； 于涛；

【作者基本信息】 天津工业大学 ， 控制工程（专业学位）， 2017， 硕士

【摘要】 无人机产生于20世纪20年代,随着高新科技的快速发展,无人机将在未来的军事以及民用领域显现越发广阔的应用前景。由于其具有尺寸小、隐蔽性好、质量轻、成本低、性价比高等优点,无人机的自主飞行关键技术备受瞩目,而无人机的定位算法是无人机自主飞行的关键技术之一。近年来,许多国内外专家对无人机的定位方法进行了深入的探究。本文主要基于meanshift目标跟踪算法提出一种对处于悬停状态中的无人机实时定位的方法。首先,介绍了无人机定位领域的基础知识和摄像机标定理论,对meanshift目标跟踪算法进行了详细介绍,由于meanshift跟踪算法的局限性,特引入Kalman滤波技术,并运用其预测和校正技术,针对目标快速移动和发生遮挡的情况,提出了一种融合Kalman的改进meanshift算法。此算法的关键点是根据已知目标模板和候选目标模板之间的颜色直方图匹配度决定算法的下一步是通过Kalman预测进行目标追踪还是选择通过融合Kalman校正技术的meanshift算法进行目标追踪,从而得到准确的跟踪轨迹。经实验验证,这种改进的meanshift算法具有较好的鲁棒性,可以实现实验的预期目标。然后,介绍了基于meanshift目标跟踪算法的无人机位置解算方法。将基于摄像机标定的定位方法与此方法形成对比,说明本方法的可行性与实用性。本方法利用改进后的meanshift目标跟踪算法提取出目标运动轨迹,并根据机载摄像头图像中任意两点在世界坐标系下的间距与相应的在图像像素坐标系下的间距之比不变的理论知识和空间距离公式来解算悬停无人机的世界坐标。最后,在实验室环境中进行定位实验,以四轴飞行器为飞行载体,验证了根据改进meanshift目标跟踪算法解算无人机世界坐标的可行性。更多还原

【Abstract】 Unmanned aerial vehicle(UAV)was invented in the 1920s.With the development of high-tech,UAV will have a broader application prospects in the field of military and civilian.Due to the advantages of its small size,good invisibility,light weight,low cost and high performance,the key technologies of UAV’s autonomous flight catch lots of attention.The positioning algorithm of UAV is one of the key technologies in UAV autonomous flight.In recent years,many experts at home and abroad make an in-depth study in positioning method of UAV.In this paper,a new approach based on meanshift target tracking algorithm for real-time positioning of hovering UAV is presented.Firstly,the basic knowledge in the positioning field of UAV and camera calibration theory are introduced.Meanshift target tracking algorithm is introduced in detail.Due to the limitation of meanshift tracking algorithm,Kalman filtering technique is specially used.Using the prediction and correction technology,aiming at the condition of fast moving and occlusion of targets,an improved meanshift algorithm combined with Kalman algorithm is proposed.The goal of this algorithm is to obtain accurate tracking trajectory by two ways.One way is that target tracking is carried on by Kalman prediction directly.The other one is that target tracking is conducted by meanshift algorithm fusing with Kalman correction method.Which begs the question,how is the improved algorithm determining the selection of the tracking ways.It is the key point of this algorithm.In this paper,according to the matching degree of color histogram between known target template and candidate target template,the next step of the algorithm,namely,one way to track,is decided.This method proposed in this study has been testified by sets of experiments to have better robustness.It can achieve the desired goal of the experiments.Secondly,a positioning method based on meanshift target tracking algorithm to calculate UAV location is introduced.To illustrate feasibility and practicality of this method,this method is compared with the positioning method based on camera calibration.Using the improved meanshift target tracking algorithm to extract target tracking trajectory,on the basis of theoretical knowledge applying to any two points of airborne camera images about ratio invariance between the distance in world coordinate system and the corresponding distance in image pixel coordinate system,and through space distance formula,the world coordinate of hovering UAV can be calculated.Finally,the positioning experiments are conducted in the laboratory,and quadrocopter is chosen to be a flying vehicle.The experiments verify the feasibility of this method that the world coordinate of UAV can be calculated by the improved meanshift target tracking algorithm.更多还原