遥感影像目标检测问题是视觉图像识别任务的重要研究内容之一，但是在船舶遥感图像中，船舶目标小且分布稀疏，使用传统的人工神经网络(ANN)进行目标检测往往会浪费大量的计算资源。脉冲神经网络(SNN)的事件驱动与低功耗特性可以极大地节省能量消耗同时解放更多的计算资源。然而SNN神经元由于其复杂动态与不可微的脉冲操作，难以正常进行训练。作为替代，将训练好的ANN转换为SNN可以有效规避这一问题。对于转换后的深层SNN，需要大量时间步长(time steps)来维持其性能。这一过程需要大量的计算资源并对产生较大的延迟，与低功耗的研究初衷相违背。本文研究了转换后SNN需要大量time steps维持模型性能的原因，并提出了新的转换方法，基于微调的逐层转换方法；考虑硬件部署的合理性，提出了泊松群编码，相比泊松编码，泊松群编码输出的脉冲序列噪声更小，对模型性能的影响更小。实验表明，微调转换方法在SAR舰船检测数据集(SSDD、AIR-SARShip)上取得与转换前模型(97.9%、79.6%)相近的性能(96.9%、70.3%)，在PASCAL VOC数据集上也获得了较好的检测性能(49.2%)，而且对于泊松群编码，time steps相同的条件下神经元数目越多，对模型性能的影响越小，时间步长较少的条件下即可获得与输入模拟频率近似的性能。本文的研究可以提升转换后SNN的性能，减少转换后SNN对time steps的需求，并为SNN的硬件部署提供了一个切实有效的输入编码方法。

Object detection in remote sensing images is one of the essential research contents of visual image recognition tasks. However, in the field of ship remote sensing images, the ship target is small and sparsely distributed, and using traditional Artificial Neural Network (ANN) for object detection often wastes a lot of computing resources. Spiking Neural Network (SNN) can be applied in this field due to its event-driven and low-power characteristics, greatly saving energy consumption and computing resources. However, it is difficult to train SNN because of the complex dynamics and non-differentiable pulse operation of SNN neurons. Instead, converting a trained ANN to SNN can effectively circumvent training difficulties. For the converted deep SNN, many time steps are often required to maintain the performance of the SNN. Unfortunately, this process requires a lot of computing resources, contrary to the original intention of low power. This paper studies the reason why a large number of time steps are required to maintain SNN model performance after conversion and proposes a new conversion method: a layer-by-layer conversion method based on fine-tuning. In addition, considering the rationality of hardware deployment, we propose Poisson group coding. Compared with Poisson coding, the output of Poisson group coding is less noisy and has less impact on model performance. The experiments in this paper show that the fine-tuning transformation method achieves a result (96.9%, 70.3%) similar to that of YOLOv3-tiny(97.9%, 79.6%) on the SAR ship detection datasets (SSDD, AIR-SARShip) and also achieves good detection performance (49.2%) on the PASCAL VOC dataset. For Poisson group coding, the more neurons there are, the less impact on model performance under the same time steps. Performance similar to the input simulation frequency can be achieved under the conditions of a few time steps. The research in this paper can improve the performance of the converted SNN, reduce the time steps requirement, and provide an effective input encoding method for the hardware deployment of the SNN.