Hierarchical 3D Indoor Map Classification Methods And Updating Mechanism

Currency is one of the important indicators to measure the performance of maps.That is judge performance of the map by checking whether the data in the map are the actual current situation.The key to improving the current situation of the map is to improve the update frequency and efficiency of the map.Objects in the indoor map are complex and moving frequently.For example,supermarkets usually change the layout of the shelves on a weekly basis,and the exhibition center changes the indoor layout on a daily basis.The exhibition center changes the indoor layout on a daily basis.Frequent scene changes and a large number of indoor scenes present challenges for indoor map updates.Meanwhile,the accompanying information about the indoor map is changed in hours or minutes.Therefore we propose a hierarchical 3D indoor map classification methods based on the nature of indoor elements,and an indoor element layering method based on improved AlexNet convolution neural network.At first,a three-dimensional indoor map update method based on the hierarchical is proposed.Traditional version updates require continuous overall replacement,resulting in high update costs.And incremental updates in a frequently changing indoor environment,a lot of incremental analysis,resulting in huge time and computing costs.Therefore,According to the mobility of indoor elements,it is divided into five layers: infrastructure layer,stability layer,active layer,POI layer and information layer.The version and incremental update methods are used for the characteristics of each layer's internal features.The method of incremental analysis is studied,and the organization structure of the data file and the algorithm for updating the operation are designed.Due to achieve reduce update time and update cost.Secondly,In order to accurately and quickly classify indoor elements,we have studied and analyzed the classification methods of objects.Due to the indoor elements are complex and often have occlusion problems.It is necessary to obtain information from local of the object.AlexNet's first convolution phase has large convolution kernel and the large moving speed.This will reduce the resolution of the feature map,Thus reduce the accuracy.Therefore,an indoor element layering method based on improved AlexNet convolution neural network is proposed.The network structure modifies AlexNet's first convolution kernel and step size to better learn the local features of the feature.And A layer of pooling is added.in order to prevent the network from over-fitting and slow convergence.This network structure is used to layer indoor elements.Finally,we verified indoor element layering method based on improved AlexNet and the update method of the hierarchical 3D indoor map.The typical scenario is used as the experimental simulation scenario.Layering with VGG-16 and AlexNet and improved AlexNet,respectively,verified the accuracy of the improved AlexNet.Analyzing of related parameters for hierarchical updates method with incremental updates and version updates.It is verified that the updating method of the hierarchical three-dimensional indoor map proposed is available and satisfies the user's requirements for the current situation.