Research On Optimal Data Selection Method In Mobile Crowdsensing

Sensing through pictures is one of the main methods of Mobile Crowdsensing(MCS).However,the redundancy of picture data will waste a lot of network resources and storage resources,so redundant screening in the data collection stage is very important.The diversity of sensing tasks,the dynamic data flow of pictures,and the high complexity of image similarity judgment algorithms make the design of image optimal data selection methods for MCS face certain challenges.Aiming at the problem of how to use image MCS to perceive heterogeneous data aggregation pre-screening with low traffic,high efficiency and high precision,this thesis optimized the image data collection process,and proposed a dynamic search algorithm for image context information clustering and image coverage optimal selection method based on heterogeneous data.The main work and innovations include:1.Aiming at the problem of near edge similarity in dynamic clustering of picture context information,this thesis optimized the data stream clustering method based on tower tree with adaptive task constraints,and proposed a dynamic search algorithm for picture context information clustering.Firstly,according to whether the tower tree clustering algorithm clusters to the existing interval,it is divided into real branches and leaves and virtual branches and leaves.The data of real branches and leaves are uploaded directly,and the virtual branches and leaves further dynamically search for the best similarity matching interval.Secondly,based on the idea of using local expansion and dynamic reduction,the similarity interval is optimized to reduce the average distance between dynamic clustering data points.Finally,the graph is clustered to the same interval Similarity filtering of slice set.Through the simulation data and the collected data experiments,it can effectively improve the clustering accuracy of context information.2.Aiming at the problem that multiple similar image data are clustered to the same range and the optimal subset is selected to cover the perceptual target after picture context information clustering,this thesis proposed an image coverage optimal selection method based on heterogeneous data.In order to reduce the number of images covered by the target,firstly,based on the heterogeneous features of the picture data perceived by MCS,the shooting direction context information is introduced in the process of image similarity feature detection;secondly,the scene of the image set covering the sensing target is defined as the MCS picture redundancy perception model;finally,the image coverage optimal selection algorithm based on greedy idea is used to solve the problem Approximate optimal subset.According to the collected data,the experiment can effectively reduce the number of pictures that participants need to upload on the premise of ensuring coverage.3.This thesis optimized the existing participatory MCS picture data collection redundancy detection process,designed heterogeneous image data collection app based on Android smartphone,and realized the image data collection process system prototype combined with the back-end server.The low dimensional heterogeneous data clustering based on the situational information(e.g.location,time,shooting angle,etc.)of photos is used to reduce or avoid the similarity detection of high-dimensional data based on image eigenvalues,and the redundant detection of picture data collection of MCS is realized with low traffic and high efficiency.