Research On Mitigating Urban Heat Island Effect In Sponge Cities Based On Satellite Telemetry Technology

Population aging is one of the most significant demographic features of today’s society.Due to the decline of physical functions of the elderly,accidental falls have become a common phenomenon in life.For the "empty nest" elderly,if they can not be treated in a timely manner after falling,it can cause serious injuries,or even endanger their lives.Falls are one of the ten leading causes of serious health injuries and deaths among the elderly,and are a major public health problem worldwide.Therefore,it is important to identify the fall behavior of the elderly through scientific technology and make timely alarm treatment,which can reduce the health injury of the elderly and reduce the medical expenses,which has high practical value.In the current research on the recognition of fall behavior of the elderly,the misjudgment of fall behavior and similar movement behavior is more prominent,and the accuracy of fall behavior recognition is low.In addition,some research methods require expensive equipment and are limited in use,making them difficult to be widely deployed and used.This dissertation aims to solve the key technical problems of flow recognition accuracy,limited computational resources,and heterogeneous sensor data fusion in human falling behavior recognition research by using relevant sensor data information obtained from smartphones and inexpensive wearable sensing devices,so as to further improve the accuracy and practicality of human fall behavior recognition.The main research work of this dissertation include the following four aspects:(1)Research on fall behavior recognition based on movement energy expenditure feature.In order not to affect the normal life behavior of the elderly,to make full use of the functions provided by the smartphones,to reduce the inconvenience caused by wearing sensor devices,and to reduce the cost of monitoring systems,the accelerometer and gyroscope integrated inside the smartphones are employed to collect the behavioral data of the elderly in their daily lives,and the threshold analysis method is used to study the human falling behavior recognition.Based on this,a three-level threshold detection algorithm for human fall behavior recognition is proposed by introducing human motion energy consumption as a new feature.The algorithm integrates the changes of human motion energy consumption,combined acceleration and body tilt angle in the process of falling,which alleviates the problem of misjudgment caused by using only the threshold information of acceleration or(and)angle change to discriminate falls and improves the recognition accuracy.The average recognition accuracy of this algorithm is verified by experiments to reach 95.42%.The APP is also devised to realize the timely detection of fall behavior and send alarms automatically.(2)Research on fall behavior recognition based on edge computing framework.In order to solve the problem that mobile devices such as smartphones has difficulty in running recognition algorithms that require large computational resources due to limited resources in the application scenario of long-term human fall behavior monitoring,a fall behavior recognition method based on the edge computing framework is proposed,using the software-defined networking(SDN)controller’s global recognition algorithm.The proposed method is based on the Bird Swarm Algorithm(BSA),which can reduce the computational overhead of smartphones by deploying computational tasks that require more computational resources on the edge server.The proposed method reduces the computing overhead of smartphones and the running time of computational tasks while ensuring stable and efficient detection in the long term.The experimental results show that the proposed method saves about 145 ms in recognition time in all four different experimental scenarios,which is about 75% time reduction.(3)Research on fall behavior recognition based on CNN-SVM model.The traditional statistical features in the study of human fall behavior recognition are generally designed and extracted by experts in the field,which have the problems of over-reliance on manual analysis,unreliability and time-consuming,making the accuracy of classification results relatively low.Convolutional neural networks have the advantage of automatic feature extraction,but convolutional neural networks are highly susceptible to overfitting on small data sets,resulting in lower accuracy of recognition.Support vector machines can achieve high accuracy recognition based on effective feature extraction.Therefore,a human falling behavior recognition model CNN-SVM is proposed,the model based on the combination of convolutional neural network and support vector machine,which solves both the problems of unreliability and time-consuming due to manually designed feature extraction and the problem of easy overfitting in deep learning models.The experimental results show that the model improves the accuracy and automation of fall behavior recognition.(4)Research on the problem of heterogeneous sensor data fusion in fall behavior recognition.The accuracy of human fall behavior recognition can be improved by using several different types of sensors.For the data fusion problem of heterogeneous sensors when processing the data collected by several different types of sensors,a multi-task joint learning data fusion algorithm model is proposed in this paper.The model achieves decision-level data fusion of heterogeneous multi-sensor data in human fall behavior recognition,and solves the problem that recognition results are not accurate enough due to a single sensor data source.The experimental results show that the model further improves the accuracy of fall behavior recognition by 98.15% at the cost of using more computational resources.