Research On Visual Positioning System Implemented By Scene Recognition Technique Based On WebGL

In recent years,there are more and more new location-based mobile Internet services.However,the commonly used positioning technologies often have problems such as limited accuracy(such as GPS positioning)or vulnerability to environmental factors(such as Wi-Fi positioning),especially in large shopping malls,railway stations,airports,parking lots and other occasions,visual positioning is relatively more advantageous.Moreover,as Web applications are very popular,it has become a new trend to study how to use Web3D-based positioning services and improve the visual positioning efficiency of the Web end through image recognition.Therefore,this paper deeply analyzed the characteristics and technical difficulties of scene recognition,starting from three aspects including feature extraction,feature matching and building a visual database,focusing on how to efficiently obtain the visual features and location information of scene images.Firstly,SIFT(Scale-invariant feature transform)descriptor was introduced to improve the accuracy and speed of image feature extraction.Secondly,a feature clustering method based on improved visual word bag method was proposed to improve the difficulty of inserting classification data when the image feature data is large.Finally,a visual database based on Keyframe was constructed to solve the problem of redundant information in traditional sampling methods.The experimental results showed that the above method can improve the efficiency of image retrieval and reduce the time of positioning while ensuring the quality of positioning.On the other hand,in order to solve the problem of scene model jam in the positioning process,this paper proposed a Web3 D rendering method of horizon hierarchical optimization.The methods were divided into pretreatment stage and operation stage.In the pre-processing stage,OBB(Oriented-bounding Box)tree was used to assist the scene model to generate spatial units,calculate the visibility of spatial units in different perspectives,and complete the elimination of invisible units,so that part of the model can be eliminated in the offline state.In the runtime stage,the relevant calculation of detail level factors was completed,and the minimumnecessary space unit required for rendering was obtained.Using the parallel computing ability of GPU(Graphics Processing Unit),the triangular sequence of necessary space unit was generated.The experimental results showed that the above method can minimize the necessary rendering time of the scene model on the premise of guaranteeing the authenticity of the scene,thus reducing the frame rate of the model rendering and ensuring the real-time performance of the scene model interaction in the positioning process.