Improved Kalman Filtering Algorithm Based On Interactive Multi-model Of Large Screen Touch Technology

At present,human-computer interaction technology based on large screens is widely used in fields such as school education interaction,business promotion,report halls,entertainment venues,etc.,which hides great commercial and social values.Human-computer interaction on large-screen touch technology is usually expressed by voice recognition technology and computer image processing technology.Among them,because the voice recognition technology has high requirements on the surrounding environment,it is easily affected by the hustle and bustle environment,so that its application range is not very large.Most image processing technologies use corresponding sensors and other related devices to capture image signals.After the capture is completed,interactive software is used to prepare the image signals captured by the sensors in advance,including image tracking of the tracked objects.Noise,gesture background separation and feature extraction,and finally positioning and tracking of moving targets.After completing a series of operations such as image processing and target tracking,the corresponding large-screen touch technology is implemented accordingly.Since the large-screen touch interaction system using the Kinect 3D sensor has the advantages of being easy to carry,simple to operate,and cost-effective,this has promoted a new development trend for large-screen touch interaction systems.However,when the environment of target movement is complex,the depth images captured by this system are often accompanied by a lot of noise.As a result,the target positioning is not accurate enough,and the touch effect is not good.This is an urgent problem to be solved.Aiming at the main problems of human-computer interaction system,this paper proposes a large-screen touch projection system based on interactive multi-model Kalman filtering improved tracking algorithm,which is not constrained by the size of the screen and can be projected on any plane and can achieve good results Gesture touch operation effect.First,keep the gesture still for 1.2 seconds,use the Kinect sensor to capture the external gesture image signal,and then feedback the gesture signal image information to the computer software for image filtering of the gesture target,gesture image segmentation,and gesture feature extraction,etc After the static gesture image,dynamic gesture tracking is performed.Because Meanshift tracking algorithm is not effective in complex gesture movement or complex background,this paper proposes an improved algorithm based on interactive multi-model Kalman filtering.Through the predictive effect of interactive multi-model Kalman filtering,the motion trajectory of the gesture is predicted and integrated Go to Meanshift tracking algorithm to achieve better gesture tracking effect and achieve high-quality large-screen human-computer interaction.The main contributions of this article are as follows:(1)A large-screen touch projection system with Kinect 3D sensor as the hardware foundation and interactive multi-model Kalman filter algorithm as the software tracking basis was independently built,and the basic composition and corresponding technical principles of the system were introduced in detail and the mapping The relationship has successfully established the interaction area of the system and realized the accurate determination of the human-computer interaction plane.(2)This article captures the gesture signal image through the Kinect sensor,and then performs image denoising,gesture segmentation and gesture feature extraction on the captured gesture image.Because some image noise interference is often mixed in the image collection processing stage,and the accuracy problem of the sensor itself makes the large screen touch projection effect not very satisfactory,it is necessary to perform noise reduction filter processing in the image processing stage.In this paper,a switching hybrid filtering algorithm is proposed,including Wiener filtering algorithm and adaptive median filtering algorithm,which performs hybrid switching filtering algorithm processing on the captured images.Through experimental comparison and verification,the filtering algorithm can effectively filter out Gaussian noise and pepper salt noise.Then,gesture segmentation processing is performed.Gesture segmentation uses skin color detection and four-way padding to segment the contour of the hand,and then uses the relative distance coefficient to extract the feature vector,which is convenient for the tracking of subsequent dynamic gestures.The screen projection touch effect is good.(3)An improved algorithm based on interactive multi-model Kalman filtering algorithm is proposed.The improved algorithm is to integrate the interactive multi-model Kalman filtering algorithm into the Meanshift tracking algorithm,which can take into account the effectiveness and accuracy of tracking object positioning.Model matching and probability update are performed to accurately track the target motion,which solves the disadvantage that the Meanshift algorithm is prone to generate systematic errors in the target tracking stage and cannot accurately predict the direction of motion in complex environments.(4)Combine the application examples to apply the improved filtering algorithm to the interactive system in this paper.In this interactive system,we can replace part of the functions of the mouse with gestures,and realize the gestures to directly control the content displayed on the large screen and demonstrate some functions.The system design tracking effect of the improved algorithm is compared with Kinect's own tracking effect,which further verifies the effectiveness of the improved algorithm.