基于四轴运动控制平台的移动目标跟踪控制研究

发布时间：2018-10-24 13:16

【摘要】：传统的运动控制主要应用于数控领域,通用性和开放性较差,多数只能够应用于特定的场合。为此,本论文基于一款开放式运动控制平台,设计视觉检测模块组成闭环控制系统,增加平台的智能性和准确性。整个运动控制系统通过视觉检测模块对移动目标的运动信息进行实时反馈,并控制四轴运动控制平台和机械手对移动目标进行追踪和抓取,以此实现智能控制。本论文中四轴运动控制平台的运动控制器部分是ARM+PCL6045B的双CPU结构。ARM作为主控CPU,不但具有高速的运行频率,还具有丰富的外设,便于整个平台功能的扩展和裁剪。PCL6045B是一款专业的运动控制DSP,是运动控制器的从CPU,能够实现单轴定位、直线/S型曲线加减速、直线/圆弧插补以及限位和回原点等功能。为了能够对目标进行抓取,在四轴运动控制平台的基础上,扩充了机械手来实现抓取动作。另外,为了便于对四轴运动控制平台进行二次开发,提高四轴运动控制平台的开放性,整个平台的软件体系结构中对相应的运动控制功能进行了API封装。为了能够实现对移动目标的准确追踪,首先需要实现对静止目标的定位和抓取,完成由像素模型坐标系到平台模型坐标系之间的转换,然后利用得到的平台坐标控制四轴运动控制平台的各个轴和机械手进行相关动作,实现对移动目标的追踪和抓取。在视觉检测部分,本论文中利用TLD算法来对移动目标进行检测。TLD算法将传统的移动目标检测算法和传统的移动目标跟踪算法相结合,能够在目标发生遮挡的情况下对目标进行重新检测,具有较高的稳定性。该算法由跟踪器、检测器、整合器以及学习器四大模块组成。最后,为了能够使运动控制平台更加平滑地实现对移动目标的追踪,增加整个运动控制系统的实用性,本论文根据运动控制平台单轴、直线插补以及圆弧插补三种运动控制方式的性能,提出具体的运动控制方案,并在此基础上,针对典型的运动控制场景,设计了一系列实验来验证整个系统在不同场景下对移动目标跟踪抓取的准确性、实时性和稳定性。
[Abstract]:The traditional motion control is mainly used in the field of numerical control. Therefore, based on an open motion control platform, this paper designs a closed loop control system based on visual detection module to increase the intelligence and accuracy of the platform. The whole motion control system can realize intelligent control by real-time feedback of moving target information through visual detection module, and control four-axis motion control platform and manipulator to track and grab moving target. In this paper, the motion controller of the four-axis motion control platform is the dual CPU structure of ARM PCL6045B. As the main control CPU, ARM not only has high speed running frequency, but also has abundant peripheral equipments. PCL6045B is a professional motion control DSP, which can realize uniaxial positioning, linear / S curve acceleration and deceleration, linear / circular arc interpolation, limit and return to origin and so on. In order to grasp the target, the manipulator is extended on the basis of the four-axis motion control platform. In addition, in order to further develop the four-axis motion control platform and improve the openness of the four-axis motion control platform, the corresponding motion control function is encapsulated by API in the software architecture of the whole platform. In order to realize the accurate tracking of moving target, it is necessary to locate and grab the static target first, and complete the transformation from pixel model coordinate system to platform model coordinate system. Then the platform coordinates are used to control each axis and manipulator of the four-axis motion control platform to track and grab the moving target. In the part of vision detection, TLD algorithm is used to detect moving target. TLD algorithm combines the traditional moving target detection algorithm with the traditional moving target tracking algorithm. It can re-detect the target in the case of occlusion, so it has high stability. The algorithm consists of four modules: tracker, detector, integrator and learner. Finally, in order to make the motion control platform realize the tracking of moving objects more smoothly, and increase the practicability of the whole motion control system, this paper according to the single axis of motion control platform, The performance of three motion control methods, linear interpolation and arc interpolation, is presented. Based on this, a typical motion control scene is proposed. A series of experiments are designed to verify the accuracy, real-time and stability of the system in different scenes.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP273

【参考文献】