This paper presents a real-time segmentation and reconstruction system that utilizes RGB-D images to generate accurate and detailed individual 3D models of objects within a captured scene. Leveraging state-of-the-art instance segmentation techniques, the system performs pixel-level segmentation on RGB-D data, effectively separating foreground objects from the background. The segmented objects are then reconstructed into distinct 3D models in a high-performance computation platform. The real-time 3D modelling can be applied across various domains, including augmented/virtual reality, interior design, urban planning, road assistance, security systems, and more. To achieve real-time performance, the paper proposes a method that effectively samples consecutive frames to reduce network load while ensuring reconstruction quality. Additionally, a multi-process SLAM pipeline is adopted for parallel 3D reconstruction, enabling efficient cutting of the clustering objects into individuals. This system employs the industry-leading framework YOLO for instance segmentation. To improve YOLO's performance and accuracy, modifications were made to resolve duplicated or false detection of similar objects, ensuring the reconstructed models align with the targets. Overall, this work establishes a robust real-time system with a significant enhancement for object segmentation and reconstruction in the indoor environment. It can potentially be extended to the outdoor scenario, opening up numerous opportunities for real-world applications.