Research And Implementation Of Human Fall Detection Based On Kinect

The commonly used human fall detection methods include a wearable devicebased detection method,an environmental signal-based detection method,and a video-image analysis-based detection method.Compared with other methods,the computer vision analysis method has the characteristics of fast analysis and small influence from external conditions.This paper studies the principle and implementation of the dynamic and static human fall detection system based on visual images.The main contents of this paper are as follows:First,the background and significance of human fall detection research are analysed,and the current status of human fall detection methods at home and abroad as well as the advantages and main problems of these methods are summarized.Introduced the Kinect development platform principle and Kinect SDK development kit,and how to use Microsoft's application program interface(API)to get bone data.Second,the built system hardware and software platform implements dynamic video fall detection based on Kinect.The principle of dynamic video fall detection is based on the dynamic data of the joints of the human body obtained by Kinect to analyse the movement speed of the human body and the change of the position of the key points of the human body to realize the falling detection of the dynamic human body.This article uses Kinect to build a hardware platform and write related software to achieve real-time dynamic human fall detection.Thirdly,aiming at the problem of fall detection based on static images,this paper proposes a detection method based on convolutional neural network for joints of human body.Based on the detected joint point information of the human body and the position information of Kinect,the human body falls down in static images.State recognition,thus completing the fall detection function.The method first uses Kinect to complete the RBG image acquisition,and passes the RGB image data to the background deep learning server for human joint analysis.After completing the human joint analysis,the image joint point information is transmitted to the front-end Windows system.The point information is combined with the location information of Kinect to finally complete the real-time judgment of the fall state of the human body in the static image.The test results show that the accuracy of the static fall detection system designed and implemented in this paper reaches more than 90%.