基于光流法算法的Visual Map快速建立方法
发布时间：2018-08-22 13:16
【摘要】：Visual Map是一个含有丰富位置信息的图像数据库,数据库中每一幅图片或图片的特征在存储时会加入相应的位置信息.室内定位的性能与Visual Map图片的数量有关.建立庞大的图片数据库能够使得定位结果更加准确,但是花费时间成本会更大.针对这个问题,本文提出了使用光流法算法来建立图片数据库Visual Map.针对光流法用于室内图像的计算会受到光线明暗不同的影响以及相机转向会产生横向偏移的问题,本文对光流法进行了改进,并使用改进后的光流法算法对摄像机采集的图像序列进行计算,得到摄像机的自身位移,从而得到每一幅图片的对应的地理位置信息.实验结果表明,利用使用光流法快速建立的Visual Map进行室内定位,误差小于1米的概率是26%,误差小于2米的概率是70%.与传统的视觉室内定位法相比,定位精度虽然略有降低,但建立图像数据库所需时间消耗大大减少.相比于视频流快速建立Visual Map方法,定位效果相当,建立Visual Map所需的设备更少,要求更加宽松.利用光流法算法快速建立Visual Map能够很好的应用于室内视觉定位系统,特别是应用于大型场所以及室内场景多变化的场所.
[Abstract]:Visual Map is an image database with abundant location information. Each picture or image feature in the database will be stored with the corresponding location information. The performance of indoor positioning is related to the number of Visual Map images. A large database of images can make the location more accurate, but it will take more time and cost. In order to solve this problem, an optical flow algorithm is proposed to build the image database Visual Map. In view of the problem that the calculation of indoor images by optical flow method will be affected by different light and dark light and the lateral deviation of camera steering will occur, the optical flow method is improved in this paper. The improved optical flow algorithm is used to calculate the sequence of images collected by the camera, and the displacement of the camera itself is obtained, and the corresponding geographic position information of each picture is obtained. The experimental results show that the probability of the error less than 1 meter is 26 and the probability of error less than 2 meters is 70. Compared with the traditional visual indoor positioning method, the positioning accuracy is slightly reduced, but the time consumption to establish the image database is greatly reduced. Compared with the fast Visual Map method of video stream, the location effect is equal, the equipment needed to establish Visual Map is less, and the requirement is more relaxed. Using the optical flow algorithm to quickly establish Visual Map can be used in indoor visual positioning system, especially in large places and places where the indoor scene changes.
【作者单位】： 哈尔滨工业大学通信技术研究所;
【基金】：国家自然科学基金(61571162) 黑龙江省自然科学基金(F2016019)
【分类号】：TP311.13;TP391.41
[Abstract]:Visual Map is an image database with abundant location information. Each picture or image feature in the database will be stored with the corresponding location information. The performance of indoor positioning is related to the number of Visual Map images. A large database of images can make the location more accurate, but it will take more time and cost. In order to solve this problem, an optical flow algorithm is proposed to build the image database Visual Map. In view of the problem that the calculation of indoor images by optical flow method will be affected by different light and dark light and the lateral deviation of camera steering will occur, the optical flow method is improved in this paper. The improved optical flow algorithm is used to calculate the sequence of images collected by the camera, and the displacement of the camera itself is obtained, and the corresponding geographic position information of each picture is obtained. The experimental results show that the probability of the error less than 1 meter is 26 and the probability of error less than 2 meters is 70. Compared with the traditional visual indoor positioning method, the positioning accuracy is slightly reduced, but the time consumption to establish the image database is greatly reduced. Compared with the fast Visual Map method of video stream, the location effect is equal, the equipment needed to establish Visual Map is less, and the requirement is more relaxed. Using the optical flow algorithm to quickly establish Visual Map can be used in indoor visual positioning system, especially in large places and places where the indoor scene changes.
【作者单位】： 哈尔滨工业大学通信技术研究所;
【基金】：国家自然科学基金(61571162) 黑龙江省自然科学基金(F2016019)
【分类号】：TP311.13;TP391.41
【相似文献】
相关期刊论文 前10条
1 孙正,郁道银,陈晓冬,徐智,黄家祥,谢洪波;基于光流法的冠状动脉造影图像序列中血管运动的估计[J];工程图学学报;2003年03期
2 韩雷;王洪庆;林隐静;;光流法在强对流天气临近预报中的应用[J];北京大学学报(自然科学版);2008年05期
3 龚大墉;汪治华;杨数强;刘岩;;基于目标的局部约束光流分析[J];重庆工学院学报(自然科学版);2009年07期
4 严强;黄增喜;曹丽萍;黄蓉刚;;光流法在机车安全行驶中的应用[J];计算机应用研究;2013年04期
5 马鹏飞;杨金孝;;基于光流法的粒子图像测速[J];科学技术与工程;2012年32期
6 张永亮;卢焕章;贺兴华;谢耀华;;基于光流预测的直线对应算法[J];信号处理;2010年05期
7 孙承志;熊田忠;吉顺平;张家海;;基于差分的光流法在目标检测跟踪中的应用[J];机床与液压;2010年14期
8 王，
本文编号：2197207
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2197207.html
