事件相机驱动的空间机械臂视觉伺服  

Visual Servo of Space Manipulators Driven by Event Camera

作　　者：李金健 胡权[1] LI Jinjian;HU Quan(Beijing Institute of Technology,Beijing 100081,China)

机构地区：[1]北京理工大学,北京100090

出　　处：《空间控制技术与应用》2023年第3期28-35,共8页Aerospace Control and Application

基　　金：国家自然科学基金资助项目(12272039和11972077)。

摘　　要：针对空间机械臂携带的传统视觉系统面临图像数据量大、测量延迟高等问题,研究了基于事件相机驱动的空间机械臂视觉伺服控制方法.事件相机是一种低延迟的新型神经形态视觉传感器,具有超高时间分辨率和高动态范围,但其产生的异步事件流数据与传统相机完全不同,需要设计新的算法以适配事件流数据.基于迭代重加权拟合设计了适用于事件流数据的快速圆形特征检测与跟踪算法;发展了适配圆形特征的空间机械臂视觉伺服方法.实验表明提出的检测与跟踪算法在实现高跟踪成功率的同时还具有远超传统视觉算法的检测速度,此外从事件相机获得的高速特征反馈能够驱动机械臂完成更加精确的伺服运动.Event cameras,a novel type of neuromorphic visual sensor,offer the high temporal resolution and dynamic range,making them suitable for use in computer vision and robotics.This article addresses the challenges of high measurement delay and large data volume in traditional visual systems used in space manipulators by exploring perception and control methods based on event cameras.Asynchronous event stream data generated by event cameras differs from that of traditional cameras,necessitating new algorithms designed specifically for event data.In this paper,a fast circular feature detection and tracking algorithm using iterative reweighting fitting is designed.Building upon this algorithm,a visual servo method adapted to circular features for manipulators is proposed.Experimental results demonstrate that the proposed detection and tracking algorithm achieves a high success rate and significantly faster detection speeds than traditional algorithms.Meanwhile,the event camera’s high-speed feature feedback facilitates more precise servo motion of the manipulators.

关 键 词：事件相机 空间机械臂 视觉伺服 特征检测与跟踪 

分 类 号：TP241[自动化与计算机技术—检测技术与自动化装置]