基于机器视觉的扁弹簧在线分类及质量检测

发布时间：2017-12-26 17:20

本文关键词：基于机器视觉的扁弹簧在线分类及质量检测　出处：《郑州大学》2017年硕士论文　论文类型：学位论文

【摘要】：扁弹簧是汽车起动机的重要组成零件之一,扁弹簧轮廓的精度直接影响了汽车起动机的使用状况。目前,扁弹簧的分类和质量检测方法仍然为人工识别,不但工人劳动强度过大,容易疲劳,检测准确率不高,而且检测效率十分低下,极大的制约了扁弹簧的生产和应用。本文提出了基于机器视觉技术的扁弹簧分类和质量检测方法,对扁弹簧图像的预处理方法、分类方法和质量检测方法进行了探索和研究,开发出了扁弹簧分类和质量检测系统。实验分析了各种预处理方法的处理效果,最终选择加权平均法作为图像的灰度化方法;选择中值滤波方法作为图像的去噪方法,达到在去除图像噪声的同时保留边缘的效果;通过寻找轮廓最小外接矩形的方法,得到扁弹簧的感兴趣区域,并使用OSTU方法对图像进行二值化处理。实验研究了基于机器视觉和DDAG SVMs的扁弹簧分类算法。对预处理图像提取黑色像素分布向量,作为特征参数向量,然后采用DDAG SVMs方法对不同型号的扁弹簧的特征向量进行训练,最后对采集到的不同型号扁弹簧测试样本进行分类测试实验。根据不同扭转角下扁弹簧横截面图像的特征,研究了扁弹簧质量检测的算法,提出两种扁弹簧质量检测方法:基于SVM的扁弹簧质量检测方法和基于图像轮廓特征的质量检测方法,并对两种方法的结果进行误差分析。最后,根据扁弹簧分类和质量检测的任务要求,设计出扁弹簧分类和质量检测视觉系统的总体方案,利用Microsoft Visual Studio、C++语言、MFC、Open CV等开发工具,完成了扁弹簧分类和质量检测系统软件的开发。研究表明,基于机器视觉的扁弹簧分类和质量检测方法效率高,准确率高,可以满足扁弹簧自动化分类和质量检测的各项要求。
[Abstract]:The flat spring is one of the important components of the automobile starter, and the precision of the flat spring contour directly affects the use of the automobile starter. At present, the classification and quality inspection methods of flat springs are still artificial identification. Not only are workers too strong in labor, but also easy to fatigue. The detection accuracy is not high, and the detection efficiency is very low, which greatly restricts the production and application of flat springs. In this paper, the classification and quality inspection method of flat springs based on machine vision technology is proposed. The preprocessing methods, classification methods and quality detection methods of flat spring images are explored and researched, and the classification and quality detection system of flat springs is developed. The experimental analysis of the effect of various pretreatment methods, the final choice of the weighted average method as gray image method; median filtering method for image denoising method, to eliminate noise while preserving the edge effect; contour by finding the minimum enclosing rectangle method, get the regional flat spring of interest. And the image binarization processing using OSTU method. The classification algorithm of flat spring based on machine vision and DDAG SVMs is studied experimentally. The distribution vector of black pixels is extracted from preprocessed image as feature parameter vector. Then, DDAG SVMs method is applied to train characteristic vectors of different types of flat springs. Finally, classification tests of different types of flat spring test samples are carried out. According to the different torsion angles under the flat spring cross section image features, on the flat spring quality detection algorithm, put forward two methods for detection of flat spring quality: Based on flat spring quality detection method and SVM detection method based on the quality of image contour features, and the results of the two methods of error analysis. Finally, according to the classification of flat spring and quality inspection mission requirements, the general design scheme of flat spring classification and quality inspection of the visual system, using Microsoft Visual Studio, C++ MFC, Open language, CV development tools, developed a flat spring classification and quality detection system software. The research shows that the classification and quality inspection method of flat springs based on machine vision is efficient and accurate, which can meet the requirements of automatic classification and quality inspection of flat springs.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U464.142;TP391.41

【相似文献】