This paper proposes a novel sharp-edged and efficient continuous semantic occupancy mapping algorithm (SEE-CSOM) for mobile robots. The main contributions are the design of the Redundant Voxel Filter Model (RVFM) and the Adaptive Kernel Length Model (AKLM) to improve the performance of the map. RVFM filters out redundant voxels with low confidence, enabling accurate object boundaries. AKLM adaptively adjusts the kernel length based on class entropy, reducing computational complexity. The multientropy kernel inference function integrates these models to generate a continuous semantic occupancy map. The algorithm is validated on indoor and outdoor datasets and a real robot platform, demonstrating significant improvements in accuracy and efficiency. SEE-CSOM addresses the challenges of overinflation and inefficiency in existing continuous semantic mapping methods by distinguishing between different types of voxels and adapting kernel length based on class entropy. The algorithm is effective in reducing redundant voxels and improving mapping efficiency, making it suitable for real-world applications. The results show that SEE-CSOM outperforms existing methods in terms of accuracy and efficiency, with the best performance in semantic mapping tasks.This paper proposes a novel sharp-edged and efficient continuous semantic occupancy mapping algorithm (SEE-CSOM) for mobile robots. The main contributions are the design of the Redundant Voxel Filter Model (RVFM) and the Adaptive Kernel Length Model (AKLM) to improve the performance of the map. RVFM filters out redundant voxels with low confidence, enabling accurate object boundaries. AKLM adaptively adjusts the kernel length based on class entropy, reducing computational complexity. The multientropy kernel inference function integrates these models to generate a continuous semantic occupancy map. The algorithm is validated on indoor and outdoor datasets and a real robot platform, demonstrating significant improvements in accuracy and efficiency. SEE-CSOM addresses the challenges of overinflation and inefficiency in existing continuous semantic mapping methods by distinguishing between different types of voxels and adapting kernel length based on class entropy. The algorithm is effective in reducing redundant voxels and improving mapping efficiency, making it suitable for real-world applications. The results show that SEE-CSOM outperforms existing methods in terms of accuracy and efficiency, with the best performance in semantic mapping tasks.