A Fast Compression Algorithm Based on Attention-Aware for 3D Point Cloud Data

In certain hazardous environments, such as deep-sea exploration and rescue missions, teleoperation can keep human operators at a safe distance while controlling robots remotely to perform tasks. Point cloud can be utilized to create a realistic environment, providing a sense of presence to teleoperators and enhancing the accuracy, safety, and efficiency of teleoperation. This paper proposes a fast compression algorithm for 3D point cloud data based on attention awareness. The core idea of this algorithm is to use attention mechanisms to guide the point cloud compression process, aiming for more efficient point cloud transmission and rendering. Specifically, we employ mixed reality devices to capture the user's attention and apply different compression strategies based on the user's focus on the scene. For regions of high user attention, high-resolution compression is utilized to preserve the details of the scene so that the operator can more clearly observe and analyze the area. For areas less concerned by users, low-resolution compression is employed to improve compression efficiency, thereby reducing processing latency and transmission bandwidth. Experimental results demonstrate that the proposed method achieves a reduction in processing delay while improving compression ratio.