Research And Implementation Of Anomaly Detection And Location Prediction System For Elderly Travel Based On Location Service

With the advent of social aging,people are increasingly under the pressure of monitoring the elderly.When the elderly go out alone,they often fall over themselves or get lost.At the same time,with the continuous advancement of positioning technology,many elderly people wear a variety of wearable devices with location services.Therefore,the detection of abnormal conditions during the travel of the elderly and the prediction of the position of the elderly through the location service is of great significance for enhancing the guardian's monitoring of the travel process of the elderly.In recent years,many researchers have done many researches on anomaly detection and position prediction,but there are few related researches on the elderly,and some studies only consider an abnormal situation or position prediction scene,which is not easy to form landing method system.Based on summarizing the previous studies,this paper divides the travel anomalies of the elderly into travel location anomalies,travel time anomalies and dwell time anomalies,and divides the elderly position prediction scene into offline position prediction and real-time position prediction,and then proposes new anomalies detection method and position prediction one.The specific work of this paper is as follows:1.For trajectory preprocessing,this paper first proposes a denoising method that combines median and averaging filters.The method can filter out large offset noise,smooth small offset noise,and has low algorithm complexity.Then,this paper proposes a trajectory segmentation method based on safe region.This method treats the elderly residence and the nearby daily area as a safe area,and retains the trajectory outside the safe area.This method can intercept effective travel trajectories and reduce the amount of data.2.In the aspect of travel anomaly detection,this paper first proposes a segmentation DTW trajectory similarity measure method,which improves the accuracy of DTW algorithm and reduces the computational complexity.Then,after clustering the trajectories,this paper proposes to obtain the contours of each trajectory class by grid growth method,and segment the contours by distance to form a grid region sequence,and extract the midpoint of each grid region to form the class according to the sequence of features of the cluster.When the travel anomaly detection is performed,the real-time trajectory is mapped into the grid region sequence,thereby detecting the line position abnormality and the dwell time abnormality,and detecting the travel time abnormality by comparing the historical travel time.The experiment proves that the accuracy of the algorithm for detecting abnormalities in the elderly is 77.8%,which is higher than other anomaly detection algorithms.3.According to the different scenarios,this paper proposes an offline location prediction method based on time matching and real-time location prediction based on time-weighted tree.The offline position prediction divides the historical trajectory of the old man into seven groups according to the day of the week,and then extracts the popular area for each group of trajectory time segments respectively,and selects the corresponding visiting area as the predicted value according to the time.The real-time position prediction method combines the grid regions of each cluster according to the point density,extracts the high-density region as the key location sequence of the cluster,and then merges the overlapping regions of the clusters to obtain the old-age travel state tree model.In the prediction,according to the real-time location of the elderly,combined with the travel time and state tree to predict the next position of the elderly.After experimentation,we can find that the prediction accuracy of offline position prediction algorithm reaches 63.5%.The accuracy of the realtime position prediction algorithm is 78.5%.4.In order to put the subject into practical use,this paper designs a prototype system for abnormality detection and position prediction for the elderly.The system includes guardian terminal software and background software to provide emergency alarm and location prediction services for the elderly.