Research On Indoor Visual Positioning Technology In Scenes With Missing Feature Point

The rapid development of location-based services has made it imperative to obtain accurate indoor locations in a low-cost and convenient way.The shielding of satellite signals and the interference of ferromagnetic materials make it difficult to guarantee the reliability of positioning results.Positioning method based on visual information has become the main technology in indoor positioning because of its low dependence on infrastructure and electromagnetic environment.However,strong background consistency in visual images,occlusions and changes in layout will cause the loss of visual background feature points,affect the accuracy of visual information matching,and even lead to positioning failure,which brings great challenges to the engineering application of indoor positioning technology.Therefore,considering the influence of visual feature distribution on image processing speed and location information acquisition accuracy,a location model based on hierarchical visual information acquisition was constructed.Aiming at the problem that global feature detection,which is widely used in visual image processing,causes high overhead and restricts engineering applications,a hierarchical feature extraction method based on consistency is proposed.By extracting image features with different information distribution of two different layers in different granularity,the feature extraction speed and accuracy are improved.In order to solve the problem that the positioning accuracy depends on the collection density of the data set,a three-dimensional minimum sum distance method was proposed to transform the positioning problem into the optimal model of the distance from one point to several straight lines in space,and the rapid convergence of the mixed error direction vector was realized.In addition,considering that a single positioning method can not meet the requirements of complex and changeable indoor environment for positioning accuracy,a three-dimensional minimum sum distance method with constraint is proposed to avoid the generation of large positioning outliers.In order to verify the feasibility of the proposed positioning method in practical engineering applications,a positioning application based on mobile terminals was developed and the positioning performance was tested in the actual indoor environment.The experimental results show that the feature extraction method can achieve the similar extraction effect compared with SURF at a faster speed.The positioning method is not limited to a single feature extraction and pose calculation method,and can achieve relatively stable positioning effect under different image acquisition poses.90% of the positioning errors are less than 0.58 m,and the average positioning errors are less than 0.32 m.In the engineering practice environment,the sub-meter indoor positioning accuracy is realized based on the terminal positioning platform.