Interaction Projection System Based On Kinect Depth Image And Its Filtering Algorithm

Machine vision based large-screen interactive devices have been widely used in teaching interaction,commercial advertising,game entertainment and other fields,and brought great commercial value and social value.Due to the late emergence of large-screen human-computer interaction technology,the research is still not mature enough.The various types of large-screen human-computer interaction devices that have already been put into use still have deficiencies.For example,the common laser based human-computer interaction system is highly accurate,but the hardware device is expensive and the system is huge and inflexible.The human-computer interaction system based on Kinect depth sensor has become a new development trend of the large-screen human-computer interaction system because of its low cost,light weight and portability,and easy promotion.However,because the Kinect hardware itself is not precise,the depth images collected by Kinect contain a lot of noise,which results in inaccurate positioning of the system.This paper takes the Kinect sensor based interactive projection system as the research object.Based on the comprehensive investigation of the development status of domestic and foreign interactive projection technology,we independently design and build a large-screen interaction projection system based on Kinect depth image.Then proposed an arbitrary switching filter algorithm and multiple information fusion tracking filter algorithm.Finally,the improved algorithm is implemented on the system platform with VENTUZ three-dimensional editing software,and the practical application of the algorithm is realized.To sum up,the main contributions of this study are as follows:(a)An arbitrary wall interactive projection system based on Kinect sensor was built independently,and its basic architecture and related technical principles were introduced.Design an interactive control software and use projection technology and dynamic capture methods to create an intelligent interaction area.(b)A hybrid filtering algorithm based on arbitrary switching rules(abbreviated as arbitrary switching filtering)is proposed.The switching system method was introduced into the field of image filtering for the first time,and it was successfully applied to an interaction projection system based on the Kinect depth image,which improved the system tracking performance.(c)An improved tracking filtering algorithm is proposed.The algorithm combines the depth information obtained by the Kinect infrared camera with the color information of Camshift and adds Extended Kalman Filter(EKF)prediction mechanism.The improved algorithm improves the accuracy and robustness of system target tracking.(d)The improved filtering algorithm is applied to the actual interactiom projection systems.An interactive demonstration of automobile structure was designed and demonstrated on the interactive projection system by using VENTUZ three-dimensional editing software.Then the system tracking effect with improved filtering algorithm is compared with Kinect's original machine learning tracking effect to further verify the effectiveness of the improved algorithm.