Joint Stable Detection And Interactive Application System Based On RGB-D

With the rapid development of artificial intelligence technology,the era of artificial intelligence is coming.To enter this new era,it is necessary to enable machines perfectly understand human movements.Therefore,research on human motion recognition is essential.As the basis of research on motion recognition,human body pose estimation has also been the focus of people's research.Its purpose is to obtain more accurate three-dimensional position information of human joint points for human motion recognition.At present,the most mature joint detection device on the market is the Kinect device.It is a wear-free somatosensory device that extracts information on human bones and joints based on color images and depth images.Although its detection method is well,it is still the same as other joint points the detection method,also faced with the problem of incorrect detection of the joint point information caused by the jitter or the occlusion of the joint point.Currently,methods such as filtering or human body module separation are mainly used to further process the collected joint point information or image information,to obtain more accurate joint point information,but the above problem is still not perfectly solved.In view of the above problems,this paper designs a detection method,that uses the image information collected by Kinect to obtain stable human joint point information.The method first uses Kinect equipment to collect the three-dimensional position information of the human bone joint points,and its corresponding color image and depth image;Second,let the color image aligned to the depth map and substituted into the Open Pose joint point detection algorithm,to obtain the two-dimensional joint point information of the human body;Third,the joint point information obtained by the two methods is aligned with the coordinate system,and the processed data is substituted into the designed bone fitting optimization algorithm,thereby obtaining more accurate three-dimensional joint information of the human body;Finally,the method is obtained the skeletal joint point model was experimentally verified.The experiment proves that it can not only improve the problem of joint point detection error caused by occlusion,but also the stability of the obtained continuous joint point model is better than that of the Kinect joint point model.This article then combines the designed joint point stability detection method with virtual reality technology to design a somatosensory interactive system for the iron smelting site museum.The main design process of the system is as fallows: First,the ancient iron smelting scene is virtually restored using the Unity3 D virtual engine;Second,the Kinect somatosensory interactive device is used to realize the virtual interaction between the tourists and the iron smelting worker model in the scene,through the designed joint point stability detection method;Third,the design adds some interactive knowledge of iron smelting related modules;Finally,the system flow is further set and optimized to complete the design of the interactive system.