Image Based Information Fingerprint Library Construction And Indoor Positioning Technology Research

With the rise of the new generation of digital intelligence applications and the development of modern information technology,more and more industries have demanded high accuracy and reliable performance for indoor positioning technology.In the existing location service research field,image-based positioning methods have become a hot research direction in the industry for their advantages such as rich information volume and convenient data collection.However,factors such as spatial limitation of indoor environment and complex content information restrict the performance of positioning methods,and there is an urgent need to research high-precision and high-stability image fingerprint positioning technology.This paper addresses four main problems:lack of spatial information association in the fingerprint database,high application cost,insufficient fingerprint de-distribution,and insufficient fine-grained localization results,and develops indoor fingerprint localization technology research based on image information.Based on the analysis of the spatial distribution law between image fingerprint features,a hierarchical fingerprint database construction algorithm with motion consistency constraint is proposed;based on extracting duplicate regions in image fingerprints,a weighted suppression algorithm for confusion regions is proposed;finally,based on fingerprint localization methods to obtain nearest neighbor fingerprint points,a deviation point localization algorithm based on fingerprint location fine-grained is proposed.The specific research work is as follows.1.To address the problems of high computational cost and lack of spatial association information of image fingerprint library,this paper proposes a hierarchical fingerprint library construction algorithm based on motion consistency constraint.Based on the analysis of the spatial distribution pattern between image fingerprint features in the same column,we model the motion consistency constraint relationship,generate motion condition functions for the key feature points of image fingerprints,and design a regression model based on the boundary conditions of motion distribution between feature point pairs to realize the feature point screening of image fingerprints in the same column range,and the obtained feature point collection is used as the class head feature of the range to form the first layer of the fingerprint database is formed,and the set of image fingerprints in the range corresponding to each class head feature is used as the second layer of the hierarchical image fingerprint database.The formation of class head features effectively introduces the spatial correlation information between fingerprint points into the fingerprint database;at the same time,the class head features of the first layer of information can support the fingerprint database to achieve fast retrieval of convergence ranges.2.To address the problems of complex indoor scenes,high content repetition,and difficulty in ensuring high differentiation of fingerprint information,this thesis proposes the confusion sub-region weighted interference suppression algorithm(CSWS)for fingerprint optimization.The confusion sub-regions in the image fingerprint are obtained through the appropriate salient region extraction algorithm,and the confusion degree is designed in the Euclidean space by combining the similarity between these sub-regions to measure the degree of influence of the confusion sub-regions on the overall fingerprint database;and on this basis,the weight coefficients positively correlated with the confusion degree are established,and the suppression of the duplicate regions in the image fingerprint database is realized by the combination of weighting.Based on ensuring the validity of image fingerprint content,the fingerprint feature differentiation is improved.3.To address the problems of insufficient utilization of image feature differences caused by location deviation and insufficient fine-grained output location results in fingerprint localization,this paper proposes a Image Feature Difference Weight and Position Deviation learning model(IFDW-PD).The algorithm designs an image feature difference weight learning network and a location deviation coefficient mapping network for the deviation point and the nearest neighbor fingerprint point,and firstly learns and extracts the image features of the fingerprint point and the deviation point through the image feature learning network with shared parameters,and uses the feature difference weight learning model to extract the image feature difference weights of the deviation point and the fingerprint point;finally,through the location deviation coefficient model,the feature difference weights are Finally,through the position deviation coefficient model,the feature difference weight is mapped to the position deviation coefficient,and a reliable correlation between the deviation point feature difference and position deviation is established to realize a fine-grained fingerprint localization algorithm that effectively responds to the deviation of deviation point position.After experimental testing and evaluation,the hierarchical fingerprint database construction algorithm proposed in this paper effectively improves the retrieval rate while ensuring the matching accuracy;the CSWS algorithm significantly improves the accuracy and stability of the positioning system in many different indoor scenarios by suppressing the interference caused by the confusion area;the average positioning error of the positioning system implemented with the IFDW-PD algorithm is 0.53m in the indoor living scenario,1.1m in the bookstore scenario,and 1.45m in the most complex indoor shopping mall environment.The average positioning error of the positioning system implemented with the IFDW-PD algorithm is 0.53m in the indoor living scenario,1.1m in the bookstore scenario,and 1.45m in the most complex indoor shopping mall environment.The algorithm effectively improves the positioning accuracy in the range of 0-2m,and reduces the long-range positioning error and improves the positioning stability.The experimental results show that the algorithm proposed in this thesis can effectively enhance the indoor location service capability and provide new ideas for the development of modern indoor positioning technology.