Design Of Augmented Reality System Based On Target Tracking And Speech Interaction

With the continuous development of computer technology,Augmented Reality(AR)has quickly come to people's vision,but most of the current augmented reality applications simply superimpose the virtual model on the real scene video screen.The virtual model keeps the same size and shape as the object in the real scene,it can't perceive the real location of the object in the scene.In this paper,target detection technology and binocular vision technology are used to collect the position change information of specific objects in real scenes into applications in time,so that virtual objects follow the movement of specific objects,which increases the authenticity and interesting of applications.In order to enhance the human-computer interaction ability of augmented reality applications,this paper uses the remote speech recognition engine to input local voice commands and upload them to the server through the network,and carry out speech recognition.The speech recognition technology is integrated into augmented reality applications,which makes human-computer interaction convenient and efficient.Aiming at target detection technology,target detection algorithm based on Deformable Part Model(DPM)is studied.The basic feature of the algorithm is the improved HOG feature,which combines the information of symbolic gradient and unsigned gradient.The algorithm simplifies appropriately,retains useful information and simplifies calculation.The feature pyramid realizes multi-scale target detection.The combined detection method of root model and component model is adopted to allow the target to have appropriate deformation.Aiming at the target position measurement,the binocular ranging technology is used.Through the process of camera imaging,lens distortion,parameter calibration and disparity matching,an improved SGBM algorithm based on edge features is proposed.The three-dimensional coordinates of the target are obtained,the communication between software platforms is constructed,and the camera driving and data transmission are realized.Aiming at human-computer interaction,besides the traditional implementation methods such as keyboard,mouse and touch screen,with the help of speech recognition server and augmented reality development platform components,the interface functions are rewritten,key classes are re-encapsulated,and the volume is detected frame by frame using the Microphone component of Unity platform to realize real-time voice command acquisition.Writing model-driven scripts to achieve different response actions of virtual models increases the interest and immersion of augmented reality applications,and provides an efficient and convenient way for participants to interact.Finally,the experimental test of the augmented reality system is carried out,and the performance and effectiveness of each part of the system are analyzed,which proves the feasibility of the above technical scheme.