Research On Indoor Robot Localization Based On Multi-vision Data Fusion

In recent years,with the advancement of science and technology as well as the improvement of living standards,people's demand for home service robots has gradually increased.The premise that home service robots can better serve humans is localization and navigation.Only when robots achieve autonomous localization and navigation can robots perceive the environment and autonomous actions better.Therefore,in order to achieve the goal of home service robot's autonomous location and navigation,the indoor localization of mobile robot is studied.In this paper,an indoor location method based on multi-vision data fusion is proposed to locate mobile robots.The algorithm is an improvement on the pHash algorithm.Using a wheeled robot assembled by our laboratory as an experimental platform,four monocular cameras are installed on four directions of the robot to collect more scene information to improve the accuracy of positioning,and using a camera towards the ceiling to calculate the rotation angle of robot.An indoor localization method based on image fingerprint information is proposed.First,use the discrete cosine transform(DCT)to compress images and convert the space signal to the frequency domain to reduce the impact of light and redundancy.Then whitening is used to remove relevance,and singular value decomposition(SVD)is used to reduce dimension.Then a global feature description vector is formed as the unique fingerprint information of the image.Next,according to the fingerprint information,the most similar image sequence can be found in the database with geo-tagged markers,and its location information can be obtained.Finally,use the phase correlation method to calculate the rotation angle and get the robot's direction information.Experimental results show that the accuracy of the proposed method can reach 5mm,it achieves better localization accuracy than other traditional approaches.Even the indoor environment is very simple,the proposed algorithm can still robustly achieved localization,which is hard to achieve by the conventional localization.