Design Of Embedded Intelligent Monitoring System Based On ARM

With the progress of society and the continuous development of science and technology,People's security awareness is also constantly improving,especially in today's increasingly serious population ageing,for the elderly living alone indoor and outdoor security needs continue to increase,so,the use of advanced video monitoring technology to monitor the elderly living alone,able to find and deal with the elderly living al-one in a timely manner,to ensure the personal safety of the elderly living alone is an urgent need of the current society.For this social problem,this thesis uses ARM embedded technology,wireless communication technology and OPENCV visual library to realize an intelligent monitoring system which integrates video monitoring,behavior recognition and abnormal notice.First of all,this thesis selects S3C4412 development board with Exynos 4412 processing chip,and proposes a design scheme of intelligent video monitoring system.In the system server construction aspect,based on the Linux operating system and uses the b/s structure to complete the entire software design;The camera is connected by USB interface and the image data is collected by V4L2 camera driver,and the H.264 code base which is transplanted on the development board is used to encode and compress the video collected by the camera Furthermore,the video resource is sent to the browser through Wi-Fi based on TCP communication protocol,so that the user can observe the environment in real time.Secondly,according to the system requirement analysis,we use the API provided by OpenCV to initially process the image and model the original image with mixed gauss background Aiming at the problem that the object recognition is not accurate enough in the process of the mixed gauss background modeling method,the processed image is fused by YOLOv3 Algorithm,the image in motion is captured in detail,and the position state and the target accuracy are guaranteed according to the character that the target size changes at any time.Then the image is tracked by the method of fusing Kalman filter and mean shift,in order to achieve the effect of High real-time and accuracy in the tracking process and complete the whole target tracking process,the feature samples are extracted by Sift Algorithm,using visual word bag and Support vector machine model,we can learn and train the image to recognize abnormal behavior.If there is an accident,the server will report the abnormal situation to the family in time.In the end,the whole function of the system is tested.The client can get the video monitoring content in the local area network.After the image is processed by the server,the client can send the alarm information to the Guardian in real time when the anomaly is detected.In the system test,the success rate of fall detection is 92.3%,which can meet the need of intelligent monitoring system and pave the way for the next research.