Spiking Neural Networks (SNNs) have attracted extensive research attention for their outstanding energy efficiency in 2D visual recognition, demonstrating increasing application potential. In current deep learning-based methods for 3D point cloud classification, the significant computational cost and energy consumption caused by numerous multiplication and accumulation operations pose challenges. While many scholars have extended SNNs to 3D point cloud recognition tasks, their classification performance remains unsatisfactory. To address this issue, the paper proposes an innovative point-to-spike classification. Firstly, it encodes the original point cloud to extract its geometric features. Then, through an image projection module, it transforms the projected point cloud data into a format suitable for image representation. Finally, by combining the spike-driven residual structure with Transformer, a spiking neural network structure is introduced for efficient classification of point cloud. This approach achieves higher accuracy in point cloud classification while maintaining low energy consumption. Through experimental validation, the proposed method achieves accuracies of 91.98% and 86.19% on the benchmark datasets ModelNet40 and ScanObjectNN, respectively. Compared to other SNN-based point cloud classification networks, the proposed method in this paper exhibits significant performance improvement.