摘要
针对室内定位问题,提出一种利用WiFi和视觉信息实现定位的算法。首先利用坐标变换原理分析并建立物体的图像平面坐标与世界坐标系坐标之间的数学关系,再以2个信标之间的距离作为约束条件建立数学方程,求解摄像机的俯仰角;然后根据俯仰角分别计算观察点与2个信标的平面距离,最后根据计算得到的距离和信标的坐标实现观察点定位。依据上述定位求解过程将会解析得到观察点的多个虚拟坐标。为此,利用SAE与DNN模型,以多个自定义WiFi特征为输入,11个距离类别为标签,依据节点之间的距离来识别定位坐标的有效值。实验结果表明,此定位算法能够在室内环境下提供高精度实时定位,算法的可靠性得到了充分验证。
To solve the indoor positioning problem,an indoor positioning algorithm based on WiFi and visual information is proposed.Firstly,the mathematical relationship between the image plane coordinates and the coordinates of the world coordinate system is analyzed and established by using the principle of coordinate transformation.Secondly,the mathematical equation is established with the distance between two beacons as the constraint condition,and the elevation angle of the camera is solved.Thirdly,the plane distance between the observation point and the two beacons is calculated according to the elevation angle.Finally,the distance between the observation point and the beacon is calculated.The marker can locate the observation point.According to the above positioning process,the virtual coordinates of the observation points will be obtained.Using SAE and DNN model and taking multiple custom WiFi features as input and 11 distance categories as labels,the valid values of positioning coordinates can be identified by the distance between nodes.The experimental results show that the algorithm can provide high precision real-time positioning in indoor environment,and the reliability of the algorithm is fully verified.
作者
贾玉福
刘文平
胡胜红
JIA Yu-fu;LIU Wen-ping;HU Sheng-hong(School of Information Management and Statistics,Hubei University of Economics,Wuhan 430205;School of Information Engineering,Hubei University of Economics,Wuhan 430205,China)
出处
《计算机工程与科学》
CSCD
北大核心
2020年第2期229-235,共7页
Computer Engineering & Science
基金
国家自然科学基金(61672213)
湖北省自然科学基金(2018CFB721,2019CFB765)
湖北省高等学校优秀中青年科技创新团队项目(T201714)
湖北省教育厅科技处研究计划(D20182202)。
关键词
单目视觉
坐标变换
室内定位
卷积神经网络
WiFi特征相似度
monocular vision
coordinate transformation
indoor positioning
convolutional neural network(CNN)
WiFi feature similarity
