巡检机器人自主仪表视觉识别系统的设计与研究

发布时间：2018-03-26 13:00

本文选题：仪表识别　切入点：物体检测　出处：《哈尔滨工业大学》2016年硕士论文

【摘要】：图像传感器作为一种非接触式高精度传感器目前正被广泛的应用于安防、医疗、工业等领域。集图像传感器、定位导航设备、声音传感器及红外热像仪为一体的巡检机器人,更是越来越多的被应用于城市安防巡检、智能小区安防巡检以及电力系统巡检中。由于电力系统的高压、高危环境使得智能巡检机器人代替人工检查电站成为必须,而对于电力系统中众多繁杂的仪表装置进行检测及监督更是必须采用高精度的视觉算法进行处理,这是实现电力系统无人化、智能化的必经之路。针对上述问题,本文采用基于视觉显著性分析的图像预处理算法缩小模板遍历区域,结合云机器人技术提高处理速度及准确率。基于绝对对偶二次曲面像变换的摄像机自标定方法使得机器人能够实时获取目标仪表空间位置信息,从而控制云台进行相应的微调,以防止目标偏移图像中心,因此调整过后摄像机变焦过程中不会丢失目标。具有流程简单、无需立体匹配,能够防止透视畸变的作用。针对室外环境下仪表外壳存在反光及雨后水珠聚光导致的仪表图像高亮区域影响识别这一特殊情况,进行了图像修复算法的研究。本文采用轮廓特征拟合规则几何图形的方法,判断指针与某一固定特征之间的位置关系,从而得到仪表的相应读数。该法对光线敏感度低,对透视畸变有一定的抑制作用。另外为搭建实验平台,本文设计了二自由度的云台机构,给出了相应的硬件平台,并基于Visual Studio 2010编写了用户操作界面,分析了系统误差,并在转动平台上进行相应实验验证。
[Abstract]:As a kind of contactless high precision sensor, image sensor is widely used in security, medical, industrial and other fields. It includes image sensor, positioning and navigation equipment, sound sensor and infrared thermal imager. It is more and more used in urban security inspection, intelligent residential security inspection and power system inspection. Because of the high voltage of power system, high risk environment makes it necessary for intelligent inspection robot to replace manual inspection power station. The detection and supervision of many complicated instrument devices in power system must be handled by high precision visual algorithm, which is the only way to realize the inhumanization and intelligentization of power system. In this paper, the image preprocessing algorithm based on visual salience analysis is used to reduce the traversal region of template. Combined with cloud robot technology, the processing speed and accuracy are improved. The camera self-calibration method based on absolute dual Quadric image transformation enables the robot to obtain the space position information of the target instrument in real time, thus controlling the cloud head to make the corresponding fine adjustment. In order to prevent the center of the target offset image, so after adjusting the camera zoom process will not lose the target. Can prevent the effect of perspective distortion. In view of the special situation that the instrument shell has reflective light and water droplets focus after rain to recognize the influence of the bright area of the instrument image in the outdoor environment, In this paper, the method of contour feature fitting rule geometry is used to judge the position relationship between pointer and a fixed feature, and the corresponding reading of the instrument is obtained. The method has low sensitivity to light. In addition, in order to set up the experimental platform, this paper designs a two-degree-of-freedom cloud head mechanism, gives the corresponding hardware platform, and compiles the user interface based on Visual Studio 2010, and analyzes the system error. The corresponding experiments were carried out on the rotating platform.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】