关键词: 空间点目标, 神经形态学, 事件流数据, 目标探测, 脉冲神经网络

Abstract:  As the onorbit spacecraft face more and more threats，how to identify and evaluate the impact of these threats on the normal operation of the spacecraft has become an urgent problem to be solved.At present，the detection and tracking of space objects mainly rely on traditional framebased visual sensors，but these sensors have shortcomings in real-time performance and data volume.Meanwhile，neuromorphological visual sensors have been widely used in the field of moving objects detection and tracking.Due to the event stream data obtained only containing information of the changing parts in the field of view and having a microsecondlevel time resolution，the speed of object detection and tracking can reach the microsecond level，while significantly reducing the amount of data to be processed.Because of these advantages of neuromorphological visual sensors，they have become the current research focus in the field of space applications.Therefore，a space point object detection method is proposed based on a threelayer spiking neural network - space point object neuromorphological detection method，which uses only event stream data to achieve the detection and tracking of space point objects.The main components include a local motion perception layer，a global motion perception layer，and an output layer，using the fractional leaky integrate and fire neurons as the basic processing units，and utilizing their adaptability to suppress hot noise in the event stream data.The method is validated using collected space point object event stream data and event stream data from public datasets.On the collected space point object event stream data，the event denoising precision and event signaltonoise ratio of the denoising filtering part can reach 0.414 and -3.036 respectively，while the total tracking time，total tracking error count，and average tracking offset of the tracking part reach 9.3952s，100，and 0.3797 respectively.The experimental results show that the space point object neuromorphological detection method can detect single or multiple fastmoving space objects from complex event stream data and can continuously track the detected space point objects.

Key words:  , space point object, neuromorphological, event stream data, object detection, spiking neural network