【作者】 刘洋；

【导师】 吕宁；

【作者基本信息】 哈尔滨理工大学 ， 控制工程（专业学位）， 2020， 硕士

【摘要】 随着人类社会由信息时代向智能时代的迈进,人工智能及相关领域的技术发展突飞猛进,作为物理信息系统重要支撑技术的机器视觉研究亦日趋活跃,平面自由运动体在交通运输、航空航天、军事、体育等方面广泛存在,对其开展实时识别、定位与轨迹预测对促进相关领域的技术进步具有重要意义。本文以“OMRON杯”Sysmac自动化控制应用设计大赛冰球对抗机器人为背景,基于运动方程建模预测方法、无监督模式循环递归网络预测方法,利用adams仿真与webots仿真研究对平面自由体的识别、定位、轨迹预测及机械手接球过程,主要研究内容如下:首先,基于小球运动特征采集实时数字的图像,利用颜色特征提取和霍夫变换等进行图像处理,对小球进行识别和定位。通过将RGB值转换为对颜色更加敏感的HSV值,再对其进行霍夫变换检测,识别出小球。避免了光照强度、角度、多背景等环境因素的干扰。然后,为了实现机械手连续稳定击球,开展对小球桌面运动轨迹预测方法研究。建立了平面自由体滑动、碰撞过程的力学模型;并基于力学模型方法、循环递归神经网络方法建立了平面自由体运动过程轨迹预测模型。为了验证所提出算法模型的准确性,提出一种Adams-Simulink联合仿真方法对算法进行验证。在此基础上,基于Webots仿真软件对平面自由体机器人系统全运动过程,包括小球桌面运动、图像识别与轨迹预测、机械手击球等进行模拟仿真。完成了webots三维模型建模,搭建还原了真实场景下的小球机器人系统,在此基础上,采用python语言编写控制器程序,包括图像采集、特征识别、建模轨迹预测等控制代码。更多还原

【Abstract】 As human society by the information age to intelligence era,in the field of artificial intelligence and related technology development by leaps and bounds,as an important supporting technology of the physical information system machine vision research also more and more widely applied,plane freedom of movement in the transportation,aerospace,military,sports,etc widely exist,to carry out real-time identification,location and track prediction is of great significance to promote technological progress in the field of related.This article takes the "OMRON cup" Sysmac automation application design contest hockey against robot as the background,through the analysis of the movement equation of compared the traditional prediction methods,unsupervised mode cycle recursion network prediction method,using Adams simulation with webots simulation realized the recognition of planar free body,positioning,prediction and the manipulator catching process,the main research content is as follows:Firstly,real-time digital images are collected based on the motion characteristics of the ball,and color feature extraction and hough change are used for image processing to identify and locate the ball.By converting the RGB value to the HSV value which is more sensitive to the color,the ball is identified by hough change detection.It avoids the interference of light intensity,Angle,multi-background and other environmental factors.Then,in order to realize the manipulator to hit the ball continuously and stably,carry out the ball desktop motion trajectory prediction method research.The mechanical model of free-body sliding and collision is established.Based on mechanical model and cyclic recursive neural network,the trajectory prediction model of planar free body motion is established.In order to verify the accuracy of the proposed algorithm model,an example of Adams-Simulink co-simulation method is proposed to verify the algorithm.On this basis,Webots simulation software was used to simulate the whole motion process of the planar free-body robot system,including ball desktop motion,image recognition and trajectory prediction,and manipulator hitting.The 3d model modeling of webots was completed,and the ball robot system in the real scene was built and restored.On this basis,the controller program was written in python language,including image acquisition,feature recognition,modeling trajectory prediction and other control codes.更多还原