Research on Interaction Optimization of Augmented Reality Technology in Visual Communication Design and Presentation

This paper proposes an enhanced ORB-SLAM3 model that integrates six-layer quadtree feature management, Farneck depth optical flow velocity compensation and second-order spherical harmonic function dynamic lighting estimation to solve the problems of spatial registration sensitivity caused by lighting and fast moving in virtual visual system, insufficient frame rate in high precision 3D asset rendering and high rate of accidental touches occurring during gesture interaction. To alleviate rendering load it uses a progressive mesh hierarchy detailing with a quadratic error metric, ASTC 6x6 hardware compression, and a Vulkan streaming loading strategy with lightweight screen-space ambient occlusion. A long short-term memory network 1SVM recognition model of temporal gesture recognition is built at the interaction layer and an adaptive touch magnification and anti-accidental touch latency mechanism is used based on dynamic Fitts’s law. The million-facet model is rendered at 60.1 fps with a 16.3ms end-to-end latency, an accidental touch rate of 2.1% and a task completion time of 8.9 seconds.