Augmented Reality (AR) aims to seamlessly integrate virtual and real-world environments, enabling immersive and interactive user experiences. Semantic segmentation technology plays a critical role in enhancing AR systems by allowing them to understand and interpret the semantics of real-world scenes. This capability ensures that virtual objects are appropriately placed within real environments, facilitating seamless interaction between virtual and physical elements. We developed an AR application leveraging the ARDK and Unity engine, incorporating Meshing and Plane Detection techniques to design a virtual object placement system with scene semantic understanding. A series of experiments were conducted to evaluate the system's performance in terms of reaction speed, computation cost, and area accuracy.