视觉定位中人工路标的属性提取算法研究

Research of Property Extraction Algorithm Based on Artificial Landmark in Vision Positioning

在基于人工路标的视觉定位中,首先进行人工路标的属性提取,构建人工路标模式,通过人工路标模式识别实现定位,因此人工路标的属性提取是定位的基础和关键。由于传感器位姿变化等原因造成人工路标图像发生旋转或轻微畸变,影响人工路标的属性计算,论文提出的属性提取算法在图像的边缘轮廓上进行,提高了定位的准确度和速度。在3m*3m平面上均匀布置了196个人工路标,每个路标由四个图块组成,图块有6种形状、4种颜色,基于VC++编写代码进行了实验验证,实验结果表明所给出的基于人工路标的视觉定位方法中图块属性计算准确,同时对于图像的旋转或形变具有不变性,能够满足视觉定位要求。

The positioning is achieved by extracting the property of artificial landmark,constructing and recognizing the artificial landmark pattern in the vision positioning based on artificial landmark,where the property extraction of artificial landmark is the base and crux.And artificial landmark image rotation and/or slightly distortion due to the changes of the sensor position-orientation or other reasons will affect the property calculation of artificial landmark,the algorithm of property calculation is based on the image contours in this paper,which will improve the accuracy and speed of the positioning.There are 196 different artificial landmarks which has four blocks layed on the 3 m*3 m plane evenly and each artificial landmark block has 6 different shapes and 4 different colors used to test the method presented.All of the code are based on VC++and OpenCV and the experimental results show that the attributions of the color block are calculated accurately,the topological pattern of the artificial landmarks construction and identification method is feasible and reliable,which can meet the requirements of monocular vision indoor positioning.