基于异步事件流的空间点目标跟踪方法

A Space Point Object Tracking Method Based on Asynchronous Event Stream

在当前的空间态势感知领域,针对快速移动的空间点目标的检测与跟踪,传统的基于帧的视觉传感器表现出了一定的局限性,难以满足日益增长的任务需求.因此,基于神经形态学的事件相机,凭借其高时间分辨率和高动态范围,已成为目前的研究焦点.提出一种基于异步事件流的空间点目标跟踪方法.通过单层脉冲神经元滤除其中的噪声,并得到候选目标;采用最近邻运动轨迹关联对候选目标进行持续跟踪,从而得到每个候选目标的运动轨迹;通过特征权重虚警滤除去除候选目标中的虚警目标,保留实际的空间点目标的运动轨迹.在实验阶段,分别使用CeleX-V事件相机测量事件数据和公共空间目标事件数据集(EBSSA数据集)验证了所提出算法的有效性.实验结果显示,在灵敏度和信息量两项指标上,相较于文中提到的基于事件的空间目标跟踪方法具有一定的优势,证明了基于异步事件流的空间点目标跟踪方法能够准确地从原始事件流数据中检测出一个或多个空间点目标,并获取其运动轨迹.

Within the current domain of space situational awareness,the traditional frame-based visual sensors have certain limitations in detection and tracking of swiftly moving space point objects,struggling to meet the escalating demands of tasks.Thus,neuromorphic event-based cameras,with their high temporal resolution and dynamic range,emerge as a focal point of research.A space point object tracking method is proposed based on asynchronous event stream.Initially,due to the substantial noise present in the raw event stream data,noise is filtered out through a single-layer spiking neuron,yielding potential objects.Subsequently,these candidates are persistently tracked using nearest neighbor motion trajectory association,which elucidates the movement trajectories of each.Ultimately,spurious candidates are eliminated through feature weight false alarm filtering,retaining the genuine motion trajectories of space point objects.During the experimental phase,the CeleX-V event camera is employed to measure event data and the public space object event dataset(EBSSA dataset)is utilized to substantiate the efficacy of the proposed algorithm.Notably,in terms of sensitivity and Informedness content,it possesses distinct advantages over the event-based space object tracking methods mentioned within the literature,affirming that the asynchronous event stream-based tracking can accurately detect one or multiple space point objects from raw event stream data and capture their motion trajectories.