单目相机物体位姿估计方法研究

现有的机器视觉通常以边缘轮廓和角点作为特征,因此要求背景单一,对环境结构化依赖程度高。为了拓展机器人的应用范围,使其脱离结构化的环境,提出了一种基于SIFT特征点和PNP技术的单目相机估计目标物体位姿的方法。以BumbleBee双目相机为硬件基础,以C++为开发平台,结合了Eigen计算库、OpenCV图像处理库和Triclops库,开发了单目视觉位姿估计算法,实现在复杂背景下对表面纹理较为丰富的物体的位姿估计。利用试验对所提方法进行了验证,试验结果表明,该算法具有较高的估计精度,可以作为机器抓取的依据。     

机器人在轨装配无标定视觉伺服对准方法

针对在轨装配过程中机器人"手眼"关系无法进行有效标定及机器人系统和被操作物惯性参数不定的情况,在传统的无标定视觉伺服基础上设计了深度估计器,基于机器人和图像运动的测量数据在线估计目标特征的深度值,并在机器人关节控制环中设计滑模控制器实时控制机器人关节运动,根据反馈图像信息纠正系统误差完成对准跟踪,通过仿真验证了方法的有效性。所提的无标定视觉伺服对准方法使机器人在装配过程中免去了复杂的"手眼"关系的标定程序,克服了机器人系统及被操作物惯性参数不确定性给装配精度造成的影响,提高了"手眼协调"的鲁棒性,保证机器人能够在复杂的太空环境下完成在轨装配任务。     

基于红外视觉/激光雷达融合的目标识别与定位方法

近年来,基于可见光图像的目标识别在无人车感知领域得到了广泛应用.然而,可见光图像目标识别无法应用于弱光和黑暗环境.针对于此,提出了一种基于红外视觉/激光雷达融合的目标识别与定位算法.首先,通过基于颜色迁移的数据增强训练方法,提高了红外目标识别算法的泛化性能.继而,提出了一种基于激光雷达修正的单目深度估计方法,通过视觉图像与激光雷达点云的数据融合,实现了基于稠密深度图对目标位置的估计,提高了对小目标的定位能力.试验结果表明,与传统Yolov3目标识别算法相比,该算法平均识别准确率可提升5.8％;此外,相对定位算法将包含小目标在内的物体相对定位精度提高了13.4％.     

Exploring 2D projection and 3D spatial information for aircraft 6D pose

The 6D pose estimation is important for the safe take-off and landing of the aircraft using a single RGB image. Due to the large scene and large depth, the exiting pose estimation methods have unstratified performance on the accuracy. To achieve precise 6D pose estimation of the aircraft, an end-to-end method using an RGB image is proposed. In the proposed method, the 2D and 3D information of the keypoints of the aircraft is used as the intermediate supervision, and 6D pose information of the aircraft in this intermediate information will be explored. Specifically, an off-the-shelf object detector is utilized to detect the Region of the Interest (RoI) of the aircraft to eliminate background distractions. The 2D projection and 3D spatial information of the pre-designed keypoints of the aircraft is predicted by the keypoint coordinate estimator (KpNet). The proposed method is trained in an end-to-end fashion. In addition, to deal with the lack of the related datasets, this paper builds the Aircraft 6D Pose dataset to train and test, which captures the take-off and landing process of three types of aircraft from 11 views. Compared with the latest Wide-Depth-Range method on this dataset, our proposed method improves the average 3D distance of model points metric (ADD) and 5° and 5 m metric by 86.8% and 30.1%, respectively. Furthermore, the proposed method gets 9.30 ms, 61.0% faster than YOLO6D with 23.86 ms.     

失效卫星远距离相对位姿估计与优化方法

针对远距离非合作慢旋目标位姿估计精度问题,提出一种融合图像超分辨与视觉SLAM的相对位姿估计方法。算法主要包含3个步骤：通过梯度引导生成式对抗超分辨技术,提升目标图像的质量以获取更多更高质量的特征点;构建特征数据库实现当前帧与特征数据库的匹配,提升旋转目标的特征跟踪稳定性;利用图优化对多帧图像进行联合位姿优化,消除累计误差,得到更为精确的估计结果。为稳定网络的训练,将自然进化算法引入到对抗训练中。为增强模型的泛化性和鲁棒性,实验中的数据集采用半物理仿真获得。实验结果表明,当等效距离为25 m且失效卫星以25（°）/s的速度旋转时,目标图像经超分辨网络增强后,能够实现连续稳定的长时间测量。     

A fast pulse phase estimation method for X-ray pulsar signals based on epoch folding

X-ray pulsar-based navigation (XPNAV) is an attractive method for autonomous deep-space navigation in the future. The pulse phase estimation is a key task in XPNAV and its accuracy directly determines the navigation accuracy. State-of-the-art pulse phase estimation techniques either suffer from poor estimation accuracy, or involve the maximization of generally non-convex object function, thus resulting in a large computational cost. In this paper, a fast pulse phase estimation method based on epoch folding is presented. The statistical properties of the observed profile obtained through epoch folding are developed. Based on this, we recognize the joint prob-ability distribution of the observed profile as the likelihood function and utilize a fast Fourier transform-based procedure to estimate the pulse phase. Computational complexity of the proposed estimator is analyzed as well. Experimental results show that the proposed estimator significantly outperforms the currently used cross-correlation (CC) and nonlinear least squares (NLS) estima-tors, while significantly reduces the computational complexity compared with NLS and maximum likelihood (ML) estimators.     

Three-line structured light vision system for non-cooperative satellites in proximity operations

Adapter ring is a commonly used component in non-cooperative satellites, which has high strength and is suitable to be recognized and grasped by the space manipulator. During proximity operations, this circle feature may be occluded by the robot arm or limited field of view. Moreover, the captured images may be underexposed when there is not enough illumination. To address these problems, this paper presents a structured light vision system with three line lasers and a monocular camera. The lasers project lines onto the surface of the satellite, and six break points are formed along both sides of the adapter ring. A closed-form solution for real-time pose estimation is given using these break points. Then, a virtual structured light platform is constructed to simulate synthetic images of the target satellite. Compared with the predefined camera parameters and relative positions, the proposed method is demonstrated to be more effective, especially at a close distance. Besides, a physical space verification system is set up to prove the effectiveness and robustness of our method under different light conditions. Experimental results indicate that it is a practical and effective method for the pose measurement of on-orbit tasks.     

空间目标地基光电探测与识别技术的发展

空间目标地基光电探测与识别技术是空间态势感知的主要技术手段之一。首先归纳了空间目标的典型光学特征,分析了用于轨道与光学特征探测与识别的两类光电望远镜特点,梳理了国外空间目标光电探测与识别技术的发展历程,简要总结了空间目标地基光电探测与识别技术和应用需求间的主要问题,展望了空间目标光电探测与识别技术的未来发展方向。     

Area-oriented coordinated trajectory planning of dual-arm space robot for capturing a tumbling target

The growing amount of space debris poses a threat to operational spacecraft and the long-term sustainability of activities in outer space. According to the orbital mechanics, an uncontrolled space object will be tumbling, bringing great challenge to capture and remove it. In this paper, a dual-arm coordinated ‘‘Area-Oriented Capture"(AOC) method is proposed to capture a non-cooperative tumbling target. Firstly, the motion equation of the tumbling target is established, based on which, the dynamic properties are analyzed. Then, the ‘‘Area-Oriented Capture"concept is presented to deal with the problem of large pose(position and attitude) deviation and tumbling motion. An area rather than fixed points/devices is taken as the object to be tracked and captured. As long as the manipulators' end-effectors move to a specified range of the objective areas(not fixed points on the target, but areas), the target satellite will be hugged by the two arms.At last, the proposed method and the traditional method(i.e. fixed-point oriented capture method)are compared and analyzed through simulation. The results show that the proposed method has larger pose tolerance and takes shorter time for capturing a tumbling target.     

Use of land＇s cooperative object to estimate UAV＇s pose for autonomous landing

The research of unmanned aerial vehicles＇（UAVs＇）autonomy navigation and landing guidance with computer vision has important signifcance.However,because of the image blurring,the position of the cooperative points cannot be obtained accurately,and the pose estimation algorithms based on the feature points have low precision.In this research,the pose estimation algorithm of UAV is proposed based on feature lines of the cooperative object for autonomous landing.This method uses the actual shape of the cooperative-target on ground and the principle of vanishing line.Roll angle is calculated from the vanishing line.Yaw angle is calculated from the location of the target in the image.Finally,the remaining extrinsic parameters are calculated by the coordinates transformation.Experimental results show that the pose estimation algorithm based on line feature has a higher precision and is more reliable than the pose estimation algorithm based on points feature.Moreover,the error of the algorithm we proposed is small enough when the UAV is near to the landing strip,and it can meet the basic requirements of UAV＇s autonomous landing.     

红外与可见光图像融合的空间碎片识别方法

针对暗弱空间环境中空间碎片的识别问题,提出了一种光照不均匀环境中的空间碎片识别方法。不同于现有识别方案,该方法从光照不均匀导致空间碎片图像源细节丢失造成识别性能下降的角度出发,首先将空间碎片的红外和可见光图像进行深度融合,并建立空间碎片融合图像数据库,然后基于训练样本采用深度学习技术训练得到空间碎片识别模型。算法分析表明,该图像融合方案具有高度的细节保留能力,识别模型具有在暗弱环境中高精度目标识别能力。最后进行了仿真实验,实验结果表明,该识别方案在姿态变化、图像源亮度变化等干扰条件下都具有较好的鲁棒性。     

基于SVR及阴影信息的SAR目标方位角估计

目标成像方位估计是合成孔径雷达（synthetic aperture radar,SAR）自动目标识别中一个重要的预处理过程。提出一种基于支持向量回归机（support vector machine for regression,SVR）并结合SAR目标阴影信息的方位角估计方法。首先通过长直边拟合法与脊波变换法对目标图像进行方位的粗估计．再由SVR完成精确估计。利用“运动与静止目标的获取与识别”（moving and stationary target acquisition and recognition, MSTAR）项目组提供的实测数据所做实验表明,此方法可以有效估计SAR目标方位角,精度高、泛化能力强．特别是在水平成像方位附近利用了目标的阴影信息,明显提高了相应区间的方位角估计精度。     

压缩感知下梯度投影在图像去噪中的应用

图像传输和接收的过程中不可避免地受到多种因素的影响而产生噪声,这样通常使得图像模糊不清难以辨识。针对去除图像高斯白噪声对图像辨识的干扰,提出利用压缩感知最优化求解思想,运用一种快速计算方法将高维数的图像信号投影在低维空间上,从而类比于压缩感知理论进行重构处理,并实现对含噪声信号的图像进行还原。实验结果证明,梯度投影法能较好地去除图像高斯白噪声,尤其对于图像中物体边界的辨识效果明显,为图像去除噪声提供了新的思路和方法。     

N-fold Bernoulli probability based adaptive fast-tracking algorithm and its application to autonomous aerial refuelling

Recently, deep learning has been widely utilized for object tracking tasks. However, deep learning encounters limits in tasks such as Autonomous Aerial Refueling (AAR), where the target object can vary substantially in size, requiring high-precision real-time performance in embedded systems. This paper presents a novel embedded adaptiveness single-object tracking framework based on an improved YOLOv4 detection approach and an n-fold Bernoulli probability theorem. First, an Asymmetric Convolutional Network (ACNet) and dense blocks are combined with the YOLOv4 architecture to detect small objects with high precision when similar objects are in the background. The prior object information, such as its location in the previous frame and its speed, is utilized to adaptively track objects of various sizes. Moreover, based on the n-fold Bernoulli probability theorem, we develop a filter that uses statistical laws to reduce the false positive rate of object tracking. To evaluate the efficiency of our algorithm, a new AAR dataset is collected, and extensive AAR detection and tracking experiments are performed. The results demonstrate that our improved detection algorithm is better than the original YOLOv4 algorithm on small and similar object detection tasks; the object tracking algorithm is better than state-of-the-art object tracking algorithms on refueling drogue tracking tasks.     

一种基于视觉成像的快速收敛的位姿测量算法及实验研究

 由三维目标与其二维图像估计相对于目标的位置和姿态是计算机视觉成像中的一个重要问题。正交迭代(OI)算法是一种快速、且能全局收敛的姿态估计方法，但是当数据恶化时，不能给出正确的旋转矩阵。本文改进了该算法中旋转矩阵求解方法，避免了旋转矩阵求解中出现的错误。建立了模拟实验系统，使用改进的算法进行测量，在0.45～5.20 m的范围内，摄像机到目标距离的相对误差小于±0.41%；在距离为3 m时，旋转角度测量误差小于±1.8°。数学仿真结果表明，改进后算法的抗噪声能力得到改善，结果更为准确，且能快速收敛。     

Adaptive fuzzy terminal sliding mode control for the free-floating space manipulator with free-swinging joint failure

Space manipulator with free-swinging joint failure simultaneously contains kinematic and dynamic coupling relationships, so it belongs to a new underactuated system. To allow the manipulator to carry on tasks, an effective robust underactuated control method for the space manipulator with free-swinging joint failure is studied in this paper. Considering the effect of model uncertainty and joint torque disturbance, a robust underactuated control system based on the Terminal Sliding Mode Controller (TSMC) is designed, but two drawbacks are discussed: (A) Robustness depraves with eliminating chattering. (B) Control parameters are difficult to be determined under unknown uncertainty and disturbance. To improve the TSMC, the adaptive fuzzy controller is introduced to estimate the real effect of unknown uncertainty and disturbance according to deviations of sliding mode and its reaching law. The estimated result is directly compensated into active joints torque. In simulation, the space manipulator with free-swinging joint executes tasks based on the TSMC and the Adaptive Fuzzy Terminal Sliding Mode Controller (AFTSMC) respectively. Same tasks can be finished with smaller joints torque and stronger robustness based on the AFTSMC. Therefore, AFTSMC can serve as an effective robust control method for the space manipulator with free-swinging joint failure under unknown model uncertainty and torque disturbance.     

一种红外多传感器对中段弹道空间邻近目标的联合超分辨弹道估计方法

 分析红外焦平面(IR FRA)对中段弹道空间邻近目标(CSO)的成像特点,指出星载红外传感器为实现对空间邻近目标的跟踪必须对其进行超分辨。提出了一种中段弹道空间邻近目标联合超分辨与弹道估计新方法。该方法结合红外焦平面成像模型和中段弹道动力学模型,使得能够同时利用红外多传感器的多帧信息,基于最小二乘准则建立联合超分辨弹道估计目标函数,并分析选择各目标的起始状态参数作为模型参数。针对目标函数的高维非线性特点,推导最小化意义下等价的降维目标函数,采用量子粒子群优化算法最优化该降维目标函数直接求解模型参数,进而计算出各目标的弹道和辐射强度,实现中段弹道空间邻近目标的联合超分辨与弹道估计。仿真结果验证了该方法的有效性,且相比于传统的先单传感器单帧超分辨、然后多传感器多帧测角数据关联与滤波方法,新方法在避免数据关联复杂问题的同时,其弹道估计精度更高、分辨能力更强。     

基于ASM实现视频中物体的定位

对目标进行有效跟踪是目前许多国家获取信息的重要手段,主动形状模型(ASM)主要应用于灰度图像中物体的跟踪与定位,本文针对Koschan等提出的结合金字塔模型的ASM方法作进一步改进,使之应用于视频(本文视频采用MPEG数字视频压缩编码标准)中物体的定位:首先将视频解码还原成图像序列,从人眼视觉特征出发,将RGB颜色模型转换为HSV颜色模型,用色调信息取代灰度信息进行物体定位;并研究图像序列帧之间物体的重合度对提高定位效率的作用.实验结果说明在ASM中使用色调信息能有效地实现视频中的物体定位.