Research On Elderly Indoor Monitoring System Based On WiFi And Binocular Vision

With the acceleration of the aging process in China,the bulky of the elderly population is gradually expanding.Because of physical function deterioration brought about by age and the lack of long-term companionship of relatives,it makes the elderly living alone face great security risks in the process of home-based care.In the event of an accident caused by a fall or even uncertain illness,it is often difficult for the elderly to take the initiative to call for help by themselves immediatelly,which is fatal to them.In order to deal with it,this paper construct the elderly indoor monitoring system which is able to issue alarms after accidents happened to keep the elderly out of danger,this system has a high theoretical value and application prospects.This paper proposes an indoor monitoring system based on WiFi and binocular vision.It uses WiFi fingerprint matching algorithm to achieve indoor positioning.It uses binocular vision to complete behavior recognition based on deep learning technology,and integrate indoor positioning and behavior identification with multiple features.To achieve a full range of indoor monitoring.Specifically,the main work is as follows:(1)An indoor location algorithm combining K-Means and WKNN is proposed.Firstly,the WiFi location fingerprint feature database is divided into different clusters by K-Means cluster analysis.Then,the closest clustering center to the WiFi data collected online is calculated according to the nearest neighbor algorithm,and finally the WKNN is passed in the cluster containing the cluster center.The algorithm calculates the position coordinates.(2)It proposes a behavior recognition algorithm based on a three-dimensional human skeleton model.Firstly,the two-dimensional human body key points are detected by MobileNet-optimized CMU_Pose network.Then two-dimensional human key points are constructed as a three-dimensional skeleton model according to binocular vision.Finally,According to the angle characteristics of the three-dimensional human skeleton model,a random forest algorithm is used to complete the behavior recognition.(3)software requirements analysis,software overall design,functional module design in according to the standardized software development process.The APP,JSP server,and PyQT development were performed separately.Finally,software testing was performed.According to the final experimental results,the WiFi indoor positioning algorithm proposed in this paper is accurate,stable and real-time.The recognition accuracy of the behavior recognition algorithm is high and stable.The elderly indoor monitoring software designed in this thesis has complete functions and strong practicability.