Robot-based Home Environment Security Status And Body Posture Recognition

The indoor environment is an important part of the safety of each person's life.Secondly,the state of the human body greatly affects the occurrence of accidents in the event of a dangerous accident.This situation is particularly prominent among the elderly population.More and more people are paying attention to the indoor safety of elderly people living alone,and indoor video security equipment has emerged.However,in the past,the method of installing the camera lacked flexibility,and usually required to install multiple cameras in the room,which required a lot of manpower and material resources.Another difficulty lies in the complex indoor environment,the occlusion situation,or the inability to capture complete human motion,resulting in inaccurate gesture recognition.With the development of camera technology,many inexpensive and powerful cameras have been born,such as Microsoft's Kinect V2 camera.Compared to traditional monocular cameras,Kinect V2 provides not only color information,but also depth information.According to the small size and flexible movement of the robot,this paper designs a home robot system equipped with Kinect V2.The home robot system designed in this paper mainly includes two maj or functions,one is to detect indoor environmental data,and the other is to identify indoor human body posture.The main work is as follows:Firstly,according to the analysis function requirements,the main structure of the robot was designed and the system configuration was carried out.Through the networking of the temperature and humidity module,the PM2.5 acquisition module,the gas detection module and the magnetic switch module,various environmental data of the room are collected.Secondly,a method for using the cumulative change of the angle of the human skeleton node vector as a feature representation under the ROS system is proposed.Using the 15-node human bone data provided by Kinect V2 and OpenNI2,the projected coordinate plane is constructed on the basis of human anatomy,and the spatial position and bone angle are calculated.The time angle pyramid method is used to encode the bone angle data of different time intervals to form a multi-level feature vector.The laboratory recorded motion video samples for multiple people and trained the SVM.The cross-validation rate of the correct recognition of the human body posture on the test set samples reached 86.95%.The experimental results show the feasibility of the human body gesture recognition method.Finally,in order to solve the drift problem when the human body is occluded,and further improve the reliability,real-time and portability of the above algorithm.The algorithm is improved by extending the Kalman filter method to estimate the coordinates of the occluded bone nodes instead of the misjudgment data in the original data.According to the adaptive energy method,the change in the skeletal angle is used as a measure of energy,and the action in the video is divided according to the equal energy.The SVM is used again to classify,and the classification accuracy rate reaches 92.1%.The test results show that the improved pose recognition algorithm has a better classification effect.Combined with the human body posture data,the home safety evaluation standard was designed.Through the integration of indoor environmental data and human body posture data,the indoor human body safety situation was better judged...