Research On Image Matching Algorithm In Indoor Visual Positioning

With the development of social informatization and smart devices,the indoor positioning research of mobile robots has become more and more demanding in the field of robots.It is also very important for the autonomous navigation of mobile robots.However,in indoor scenes,the global positioning system(GPS)cannot be effectively positioned due to the obstruction of the wall and other factors.The positioning technologies that can be used indoors such as broadband and WiFi also have problems with base station laying,line-of-sight space,and delay.Because of the simple equipment required for visual positioning technology,the development of camera hardware is relatively good and the cost is not high.Compared with other sensor cameras,the impact of environmental changes is small,so visual positioning has received widespread attention.Image matching is the most critical link in visual positioning.The effects of image matching accuracy,rate,and robustness will directly determine the results of visual positioning.Therefore,image matching has become an important part of this study.First,take a checkerboard image of different angles,and then calibrate the camera to get the internal and external parameters of the camera.Compared with several current calibration methods,the linear calibration method does not consider the distortion of the camera,so although the running speed is fast,the calibration accuracy is low.Although the non-linear calibration method has high calibration accuracy,it requires a large amount of complicated calculation.Zhang Zhengyou's calibration method only needs to use a checkerboard to achieve the calibration of the camera,which has improved in accuracy and simplicity.Therefore,in this paper,Zhang Zhengyou calibration method in the two-step method is selected.In the image matching algorithm part,this paper uses the feature point detection algorithm of the ORB algorithm to improve the AKAZE algorithm.In the feature point detection part,the oFAST algorithm is used,and then combined with the M-LDB algorithm in the AKAZE algorithm to construct the feature descriptor,and the Hamming distance method is used for image matching.Finally,the RANSAC algorithm is used to eliminate the false matching,and the matching result is obtained.After a large number of experimental comparisons,the improved algorithm has a higher matching accuracy than the ORB algorithm.Matching speed is faster than AKAZE algorithm.And the improved algorithm has good image matching performance under different blur levels,different JPEG image compression,different lighting levels and different rotation angle changes.Finally,collect indoor environment pictures to extract information to build an environmental feature database,and use a combination of global feature descriptors and local feature descriptors to search for similar images,and use the epipolar geometry algorithm to calculate similar images to determine the position of the moving carrier.Positioning results.The improved AKAZE,AKAZE and ORB image matching algorithms are used to compare the positioning results.The algorithm of this paper performs well in terms of positioning speed and accuracy.Therefore,the algorithm of this paper is suitable for image positioning systems.