新型单定子二自由度超声电机及其驱动的智能云台系统的关键技术研究

Because the traditional PTZ video surveillance system has arc-shaped two-DOF of rotational motion, and the depth of target is unknow, nonlinear feedback control is difficult. It is unable to meet the growing demand for security of Things planning, safe city and border security. The key bottleneck problem is high positioning accuracy and miniaturization of PTZ. A single stator and two-DOF ultrasonic motor is proposed for driving the PTZ. Using image matching to obtain the target center deviation from the video center, and then establishing the dual quaternion coordinate relations of previous frame image, the current frame image and PTZ, finally backstepping the changes in PTZ position and orientation, so as to achieve the precise positioning feedback control. At the same time, using the nonlinear filtering to predict the targe location of the next frame image, so as to improve the dynamic response and positioning accuracy of PTZ. The controllability and power outages since the lock of the ultrasonic motor combine image technology in the area of hardware and software, and they form the complementary. It breaks through the system barriers between the traditional motor, PTZ and image matching. The goal of the subject image acquisition has become information of the closed-loop feedback control to the motor and PTZ, so as to dramatically reduce the complexity of the hardware. It can provide a theoretical basis and technical support to independent research and development a miniaturized, high precision and high dynamic performance of the PTZ, and can enhance the capacity of the source of innovation in related fields in China.

传统的云台类视频监控系统存在呈弧线形二自由度旋转运动的目标深度不明，非线性反馈控制难的问题，已无法满足安防物联网规划以及平安城市、边境安全等日益增长的需求。云台系统的高定位精度及小型化的要求成为其亟待解决的关键瓶颈问题。提出针对云台类视频监控系统的单定子二自由度超声电机；并利用图像匹配获取目标中心距视频中心偏差，建立前一帧与当前帧的图像和云台的对偶四元数坐标关系，反推云台位姿变化，从而达到精确定位反馈控制的目的；同时，采用非线性滤波预测下一帧图像目标位置，提高云台动态响应性能及定位精度。本申请将超声电机的可控性、断电自锁性与图像技术软硬结合，形成有机互补；并将突破传统的电机、云台及图像匹配之间系统与系统的壁垒，使图像获取的目标标的又成为电机与云台运动控制的闭环反馈信息，大幅降低硬件的复杂性。为自主研发小型化、高精度、高动态性能该类云台提供理论依据与技术支持，提升我国相关领域的源头创新能力。

传统的电磁电机驱动的云台已经渐渐无法满足航空航天、安防监控、机器人等领域的应用需求，开发高定位精度、高动态性能、小型化的智能云台系统是突破瓶颈所亟待解决的问题。本项目针对传统多自由度云台需要多个电机驱动以致大而笨重的弊端，以具有诸多优点的超声电机技术作为突破口，利用单个多自由度的超声电机实现对云台绕两轴的直接驱动。提出了单定子塔形超声电机利用局部弯曲振动作为工作模态，利用摩擦力驱动负载转动。在对振子进行有限元分析和测振实验测定样机基本特性后，对云台系统整体建立机电耦合模型并进行了稳态输出特性的仿真。针对超声电机驱动云台系统中存在的参数时变、死区等非线性因素，选用模糊PID控制算法来提高控制效果。引入图像技术实现云台系统闭环控制效果。一方面，仅需要利用相机标定和基于SURF特征点匹配技术就实现了云台转角及转速输出的测定，省去了额外的位置传感器。另一方面针对位姿估算速度较慢的缺点，提出采用基于图像的视觉跟踪方法进行闭环控制，对核相关滤波器跟踪算法进行改进，得到了利用特征融合及考虑了尺度的改进目标跟踪算法。实验结果显示改进算法实现了实时性和鲁棒性的双重平衡，更具有实用性。本项目的研究表明利用超声电机进行驱动和图像技术闭环反馈来构建轻小型的智能云台系统方案是可行的，为后续研究提供理论依据与技术支持。