The paper introduces M-detector, a novel method for moving event detection using LiDAR point streams. Traditional methods, such as event cameras, achieve this task with microsecond latency but are limited to frame-level processing, resulting in tens to hundreds of milliseconds of latency. M-detector, designed based on occlusion principles, detects moving events point-by-point, achieving a latency of just 2-4 μs per point. This method leverages the high-rate sampling of ToF LiDAR sensors, eliminating the need for frame accumulation and providing superior accuracy, computational efficiency, and generalization. Experiments on various datasets and applications, including autonomous driving, UAV obstacle avoidance, traffic monitoring, surveillance, and mapping, demonstrate the effectiveness of M-detector. The method's low latency and real-time performance make it suitable for dynamic environments, offering timely and robust moving event detection with minimal processing delay.The paper introduces M-detector, a novel method for moving event detection using LiDAR point streams. Traditional methods, such as event cameras, achieve this task with microsecond latency but are limited to frame-level processing, resulting in tens to hundreds of milliseconds of latency. M-detector, designed based on occlusion principles, detects moving events point-by-point, achieving a latency of just 2-4 μs per point. This method leverages the high-rate sampling of ToF LiDAR sensors, eliminating the need for frame accumulation and providing superior accuracy, computational efficiency, and generalization. Experiments on various datasets and applications, including autonomous driving, UAV obstacle avoidance, traffic monitoring, surveillance, and mapping, demonstrate the effectiveness of M-detector. The method's low latency and real-time performance make it suitable for dynamic environments, offering timely and robust moving event detection with minimal processing delay.