节点文献

多传感器融合的高精度无缝定位模型与方法研究

Research on High Precision Seamless Positioning Model and Method Based on Multi-sensor Fusion

【作者】 刘飞

【导师】 张继贤; 王坚;

【作者基本信息】 中国矿业大学 , 摄影测量与遥感, 2020, 博士

【摘要】 室内外无缝定位信息是关系到国防安全、经济发展、社会民生的重要数据基础,发挥着重要的支撑作用,在国内外得到极大的重视和发展。然而,针对城市复杂环境以及室内等GNSS盲环境的高精度无缝定位技术仍然存在着定位模型与方法研究不足的情况。因此,本文围绕室内外无缝定位主题,开展了基于GNSS、UWB、INS和视觉等传感器融合的室内外定位模型、方法研究,形成了以下创新成果:(1)针对定位盲环境或者应急情况下存在定位基准缺失和破坏等问题,提出了一种基于UWB/GNSS技术的覆盖区域、现场和室内的三级室内外无缝定位基准构建方法,重点研究了基于UWB自组网定位基准构建方法,解决了GNSS室外定位基准向室内传递问题,实现了0.35m量级定位精度的UWB室内定位基站网络自动构建,可以保障定位盲环境且兼顾应急情况下的室内外定位基准需求。(2)针对高楼林立街道、立体交通、长距离隧道、树洞街道等复杂的城市环境对城市高精度、高连续和高可靠性导航定位的挑战,提出了一种零速约束的GNSS/INS/Odometer紧组合定位模型,通过GNSS抑制INS设备的误差漂移,利用INS辅助提高GNSS模糊度解算成功率,通过Odometer辅助解决GNSS长时间失锁时INS定位发散的问题,通过紧组合模型能够充分发挥三者互相补充的优势,实现在上述场景下高精度定位。结果表明除了长隧道环境在2分钟内,可以保持优于2米的定位精度外,其他场景基本可以实现0.1米的定位精度。(3)针对UWB测距受到室内非视距、多路径等影响测距和定位精度等情况,首先,研究了基于RBF神经网络算法的UWB测距误差改正模型,实现了UWB室内测距精度优于0.08m;研究分析了基于TOA原理的UWB定位模型,并在此基础上提出了基于改进抗差EKF的UWB室内定位模型,根据预测残差调整增益矩阵的大小,减弱或者消除了粗差对状态向量的影响。该方法平面定位中误差为0.13m,对比基于最小二乘和EKF算法,分别提高了88.98%和53.57%。(4)提出了一种UWB、PDR和地图融合高精度室内定位模型,可以通过UWB为PDR定位提供空间基准、抑制PDR定位精度发散;利用PDR提高UWB的定位频率、解决UWB信号覆盖差或者无覆盖区域的定位问题;且通过室内地图帮助抑制PDR航向角发散,以及定位结果发散等问题,相互取长补短,实现具备绝对定位基准、高频率、高精度室内定位。(5)针对非视距环境下UWB室内高精度定位问题,提出了一种基于CKF算法的UWB/INS融合定位模型,在室内非视距环境下定位,可以消除由于UWB信号受到遮挡而产生的多路径和非视距效应,而且可以增加定位结果的高频姿态信息。当IMU积分数据的误差增大时,通过UWB定位数据可以对INS定位结果进行约束。(6)针对单目视觉SLAM存在着尺度漂移、由于环境因素影响频繁初始化,导致定位不连续等问题,提出了一种顾及尺度因子的UWB/视觉融合定位模型,通过UWB与视觉的融合,可以充分发挥二者之间的互补特性,解决视觉初始化、尺度模糊和绝对空间基准等问题,提高UWB定位精度和定位频率以及减少基站的数量。该模型可以可靠地在纹理稀疏或者光线频繁变化室内环境实现0.2m量级的定位精度。(7)提出了一种编码图形辅助的单像室内定位模型,可以以单幅影像为基础,通过编码图形物方点、像方点和投影中心共线的原理,从该影像所覆盖范围内的编码图形的已知地面坐标和相应点的像坐标量测值出发,解算出摄像机在摄影时刻所处的位置。计算过程中涉及到的编码影像、编码图形、编码图形的像方和物方坐标均可以通过计算机自动识别和获取。另外,通过Tukey权因子模型,可以检测编码图形像方坐标存在的误差,并根据观测值残差大小调整参与计算的权重,进而减弱或者抑制观测值误差对定位结果的影响,实现优于0.1米量级定位精度。本文提出的理论模型和方法的可行性、可靠性和定位精度经过了试验验证,相关的模型和方法可以用于室内外行人、车辆等高精度定位。该论文有图110幅,表25个,参考文献163篇。

【Abstract】 Seamless indoor and outdoor positioning information has received great attention and development at home and abroad,which plays an important supporting role in national defense security,economic development,social and people’s livelihood.However,there is still a lack of research on location models and methods for seamless location technology in urban complex environment and indoor GNSS blind environment.Therefore,this paper has carried out the research on indoor and outdoor positioning models and methods based on the fusion of GNSS,UWB,INS and vision sensors focusing on the theme of seamless indoor and outdoor positioning,and some innovations are as follows:(1)In order to solve the problem of lack or destruction of location datum,in the case of blind location environment or emergency situation,a construction method of seamless positioning datum based on UWB/GNSS technology is proposed,which involves three-level for area coverage,on site and indoor seamless positioning datum.This paper mainly researched the construction method of positioning datum based on UWB ad hoc network,and solved the problem of transmission of GNSS outdoor positioning datum to indoor,and finally achieved the automatic construction of UWB indoor positioning base station network with positioning accuracy up to 0.35 m.It can guarantee the indoor and outdoor positioning reference requirements in the location shaded environment or emergency situation.(2)Aiming at the challenge of high-precision,high-continuity and high-reliability navigation and positioning in the complex urban environment,such as high-rise streets,three-dimensional traffic,long-distance tunnels,and wooded streets and so on,a zerospeed constrained GNSS/INS/Odometer tight integrated positioning model is proposed.The error drift of INS equipment is suppressed by GNSS,and the success rate of GNSS ambiguity resolution is improved by INS.Odometer is used to solve the problem of divergence of INS positioning when GNSS is out of lock for a long time.Through the tight combination of the three,we can give full play to the advantages of mutual benefit and achieve high-precision positioning in the above scenarios.Through the tight combination of the three,we can give full play to the advantages of mutual benefit and achieve high-precision positioning in the above scenarios.The results have shown that the model can achieve positioning accuracy of 0.1m except in the situation of long tunnel,which can achieve positioning accuracy of 2m.(3)In view of the fact that UWB ranging accuracy is affected by indoor non-line-of-sight and multi-path,firstly,the UWB ranging error correction model of RBF neural network algorithm is studied.The results have shown that the algorithm can make indoor accuracy of UWB up to 0.08 m..The UWB positioning model based on TOA principle is studied and analyzed,and on this basis,an indoor UWB positioning model based on improved robust EKF is proposed,which adjusts the gain matrix according to the prediction residual to weaken or eliminate the influence of gross error on the state vector.The plane positioning error of this method is 0.13 m,which is improved by 88.98% and 53.57% respectively compared with the least square and EKF algorithms.(4)A high precision indoor positioning model based on the fusion of UWB,PDR and map is proposed.UWB can be used to provide spatial reference for PDR positioning and restrain the divergence of PDR positioning accuracy.PDR is used to improve the positioning frequency of UWB,and the problem of poor UWB signal coverage or no coverage area can be solved at the same time.The indoor map is used to help restrain the divergence of PDR heading angle and the divergence of positioning results.The three learn from each other to achieve absolute positioning datum,highfrequency,high-precision indoor positioning.(5)Aiming at the problem of UWB indoor high-precision positioning in non-lineof-sight environment,a UWB/INS fusion positioning model based on CKF algorithm is proposed.Positioning in indoor non-line-of-sight environment can eliminate the multi-path and non-line-of-sight effects caused by occlusion of UWB signals,and increase the high-frequency attitude information of the positioning results.When the error of IMU integral data increases,the INS positioning result can be constrained by UWB positioning data.(6)Aiming at the problems of monocular vision SLAM,such as scale drift,discontinuous location caused by frequent initialization due to environmental factors,a UWB/ vision fusion localization model considering scale factor is proposed.Through the integration of UWB and vision,we can give full play to the complementary characteristics of the two,and solve the problems of visual initialization,scale blur and absolute spatial reference,and improve the positioning accuracy and frequency of UWB,and reduce the number of base stations.The model can reliably achieve a positioning accuracy of 0.2m in indoor environments with sparse texture or frequently changing light.(7)In this paper,a single image indoor location model aided by coding graphics is proposed.On the basis of a single image,through the principle of collinear between the object point,the image point and the projection center of the camera,and from the known ground coordinates of the coding graphics and the measured values of the image coordinates of the corresponding points in the area covered by the image,the position of the camera at the time of photography is calculated.The coded images,coded graphics and the image and object coordinates of the coded graphics involved in the calculation process can be automatically identified and obtained by the computer.In addition,through the Tukey weight factor model,we can detect the error existing in the coordinate of the coding image,and adjust the weight to participate in the calculation according to the residual error,so as to reduce or restrain the influence of the error of the observation on the positioning result,and the positioning accuracy is better than 0.1m.The feasibility,reliability and positioning accuracy of the theoretical models and methods proposed in this paper have been verified by experiments.The relevant models and methods can be used for indoor and outdoor pedestrians,vehicles and other highprecision positioning.There are 110 figures,25 tables,and 163 references in this dissertation.

【关键词】 无缝定位GNSSUWB视觉INS
【Key words】 Seamless positioningGNSSUWBvisionINS
节点文献中: 

本文链接的文献网络图示:

本文的引文网络