基于智能机器人的仪表示数识别技术与系统研制

发布时间：2018-02-28 04:35

本文关键词： 仪表识别结构化支持向量机 SURF特征预建模 KNN　出处：《浙江大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着科学技术的发展,越来越多的领域开始使用智能系统。仪表识别技术作为一种智能处理技术,已被广泛地应用在工业领域上,且越来越受到人们的关注。虽然国内外学者对固定式仪表识别已经有了大量的研究,但是对于非固定式仪表识别的研究还非常少。本文研发了基于智能机器人的非固定式仪表示数识别系统,它通过机器人搭载的摄像头获得仪表图像,并实现多类仪表的示数识别。本文主要工作包括仪表识别的算法研究和面向电力领域中非固定仪表识别的多类别仪表示数识别系统的研发。为解决非固定仪表识别中存在的仪表位置的随机性问题,同时降低仪表识别难度,提高识别率,本文提出了基于仪表检测、仪表配准和仪表识别的计算模型。对于多类型的仪表识别难题,本文设计了基于仪表建模的识别方法,降低了识别算法的复杂性,提高了识别算法的精度和鲁棒性,并实现了系统的通用性和兼容性。同时本文还对仪表识别算法中的若干方法做了相应的改进。本文首先提出了一种基于结构化支持向量机的仪表检测算法。该算法使用结构化支持向量机作为分类器,充分利用了目标与背景中其他物体的几何关系,提高了检测准确率。同时算法使用限制对比度自适应直方图均衡的方法进行图像预处理,有利于减少光照等环境因素对图像的影响。其次,本文提出了基于SURF特征的图像配准算法。该算法通过SURF特征和BF算法进行特征匹配,使用PROSAC算法进行匹配对筛选。同时本文在特征点匹配算法上做了一定的改进,包括限制配准区域和对匹配对的预筛选。图像配准算法是后续识别算法的依赖,对降低识别难度,提高识别精确度有较大的帮助。接着,本文针对两类常用的仪表类型提出了识别算法。其中指针式仪表识别基于预建模算法和图像旋转法。预建模算法充分利用了仪表模板的先验信息,降低了指针检测的难度。图像旋转法通过旋转图像后在水平区域提取匹配样本,相比直接旋转搜索窗口,降低了难度和计算量。数显式仪表识别算法通过KNN分类器识别数字。该算法充分利用先验信息进行数显区域的倾斜矫正和数字粗略定位,提高了数字分割的准确性。算法同时提出了使用轮廓和凸包的关系单独识别小数点的方法,相比直接将小数点进行分类识别有更好的效果。最后,针对与大立公司的合作项目,本文研制了基于智能机器人的仪表示数识别系统,该系统由仪表建模软件和仪表识别算法组成,完整地实现了智能机器人从预置点停下后进行检测、配准、识别的整个过程,具有较高的准确率和速度。该仪表示数识别系统已得到实际应用。
[Abstract]:With the development of science and technology, more and more fields begin to use intelligent system. As an intelligent processing technology, instrument recognition technology has been widely used in the industrial field. And more and more people pay attention to it. Although scholars at home and abroad have done a lot of research on fixed instrument recognition, However, there are few researches on the recognition of non-stationary instruments. In this paper, an intelligent robot based non-stationary instrument representation recognition system is developed, which obtains the instrument image through the camera of the robot. The main work of this paper includes the algorithm research of instrument recognition and the research and development of multi-class instrument representation number recognition system for non-fixed instrument recognition in electric power field. In order to solve the problem of non-fixed instrument recognition. The randomness of the location of the instrument, At the same time, the difficulty of instrument identification is reduced and the recognition rate is improved. This paper presents a calculation model based on instrument detection, instrument registration and instrument recognition. The complexity of the recognition algorithm is reduced, and the accuracy and robustness of the recognition algorithm are improved. At the same time, some methods of instrument recognition algorithm are improved. Firstly, a new instrument detection algorithm based on structured support vector machine is proposed. Method using structured support vector machine as classifier, The geometric relationship between the object and other objects in the background is fully utilized, and the detection accuracy is improved. At the same time, the algorithm uses the method of constrained contrast adaptive histogram equalization for image preprocessing. It is helpful to reduce the influence of environmental factors such as illumination on the image. Secondly, an image registration algorithm based on SURF features is proposed. The algorithm uses SURF feature and BF algorithm to match the image. At the same time, this paper makes some improvements in the feature point matching algorithm, including limiting the registration region and pre-screening matching pairs. Image registration algorithm is dependent on the subsequent recognition algorithm and reduces the difficulty of recognition. It is helpful to improve the accuracy of recognition. Then, In this paper, a recognition algorithm is proposed for two kinds of instrument types, which are based on pre-modeling algorithm and image rotation method. The pre-modeling algorithm makes full use of the prior information of the instrument template. It reduces the difficulty of pointer detection. The image rotation method extracts matching samples in the horizontal region after rotating the image, compared with the direct rotation search window. The digital display instrument recognition algorithm uses KNN classifier to recognize the number. The algorithm makes full use of prior information to correct the tilt of the digital display region and locate the digital roughly. At the same time, the method of using the relationship between contour and convex hull to identify decimal points separately is proposed, which is more effective than the classification and recognition of decimal points directly. Finally, for the cooperative project with Dali Company, In this paper, an instrument representation recognition system based on intelligent robot is developed. The system is composed of instrument modeling software and instrument recognition algorithm. The whole process of intelligent robot detection, registration and recognition after stopping from preset point is realized. The system has been applied in practice.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【参考文献】