论文总字数:26303字
摘 要
随着我国汽车产业的高速发展,轮毂生产流水线日均产量急剧增加,如此规模的生产也给后续的轮毂装配包装带来了巨大的压力,因此经常发生轮毂混包,原本一个包装内的6个同一型号的轮毂混入1个或2个其他型号的轮毂,导致该包轮毂无法交付给组装车间。而传统的人工分拣,由于工人的劳动程度大,很容易产生视觉疲劳,加之轮毂生产线的速度快等各方面原因,导致对特定型号包装内的错误轮毂识别效率很低,急需一种准确高效的自动检测方法来提高生产效率。
本文通过查阅国内外诸多文献资料结合实际轮毂生产企业提供的六种型号的轮毂图像对轮毂混包检测系统的各个环节进行了深入的研究。
首先,对实际采集到的轮毂图像进行预处理,包括灰度化,直方图均衡化,图像滤波,二值化等,增强图像对比度,消除噪声。然后,提取轮毂图像轮廓圆半径,孔洞面积比,Hu矩等特征。最后,根据第二步提取到的特征使用支持向量机分类器对采集到的图像进行分类,从而检测出混包的轮毂。通过对本系统最后的性能测试得出本文设计的算法和方案可行,检测效果可以达到生产要求。
关键词:轮毂,混包,图像预处理,特征提取,分类识别
Abstract
With the rapid development of the automobile industry, the daily output of the wheel hub production line has increasing sharply. Such large production scale brings huge pression to later packaging of the wheel hug, so immiscible packaging often takes place. In the original package, the six hubcaps of the same type were mixed into one or two other hubcaps. The package hub cannot be delivered to the assembly shop. And traditional manual sorting, due to the worker's labor intensity big, it is easy to produce visual fatigue, coupled with a wheel production line speed and so on various aspects reason, led to the incorrect hub within the specific model packaging on identification 2efficiency is low, be badly in need of an accurate and efficient method of automatically detecting the to improve production efficiency.
In this paper, many literatures at home and abroad combined with the actual wheel production enterprises with six types of wheel hub image in hub mix to the package test system carried on the thorough research.
First of all, to the actual collection to the hub of image preprocessing, including gray, histogram equalization, image filtering, binarization and so on, to enhance image contrast, eliminate noise. Then, the contour radius of the hub image is extracted, the area ratio of the hole area ratio, the Hu moment and other characteristics are the basis for identifying the classification. Finally, according to the characteristics of the second step, the classifier of the support vector machine is used to classify the collected images to detect the hub of the mix.
KEY WORDS: wheel hub, mixed packaging, image preprocessing, feature extraction, classification.
目 录
摘要………………………………………………………………………………….I
Abstract…………………………………………………………………………….II
第一章 绪论
1.1 项目研究背景与意义………………………………………………………1
1.2 国内外研究发展现状………………………………………………………1
1.3 机器视觉简介………………………………………………………………2
1.4 主要研究内容和具体工作…………………………………………………3
第二章 轮毂图像预处理……………………………………………………………5
2.1 噪声分析……………………………………………………………………5
2.1.1 噪声的来源………………………………………………………….5
2.1.2 图像噪声的模型…………………………………………………….5
2.2 图像灰度化…………………………………………………………………7
2.3 图像直方图均衡化…………………………………………………………7
2.4 图像滤波……………………………………………………………………9
2.4.1中值滤波………………………………………….…………………9
2.4.2双边滤波..............................................10
2.5 图像二值化……………………………………………… ………………11
2.6 本章小结…………………………………………………………………12
第三章 轮毂图像特征提取………………………………………………………..13
3.1 轮廓圆半径.................................................13
3.2 孔洞面积比……………………………………………………… ………15
3.4 Hu矩……………………………………………… ………………………16
3.5 本章小结...................................................18
第四章 轮毂图像分类................................................19
4.1模式识别……………………………………………………………………19
4.1.1模式识别的定义……………………………………….………….19
4.1.2模式识别与机器学习………………………………….………….19
4.2 支持向量机…………………………………………………… …………20
4.3分类效果测试………………………………………………………………22
4.4混包检测效果测试…………………………………………………………23
4.5 本章小结……………………………………………………… …………26
第五章 结论与展望..................................................27
5.1 结论.......................................................27
5.2 展望.......................................................27
致谢………………………………………………………………………………….28
参考文献...........................................................29
- 绪论
1.1项目研究背景与意义
进入21世纪以来,随着经济全球化的蔓延,人们生产力和生活水平普遍提高,全球汽车工业得到了迅猛地发展。日常生活中人们越来越离不开汽车,据不完全统计,近年来汽车的需求量急剧上升。随着新能源汽车技术的进步,未来汽车行业必将得到长足发展。轮毂作为汽车最重要的组成部件之一,起着稳定车身,前进与转向等至关重要的作用。因此在规模日益扩大的轮毂生产过程中,既要提高生产效率,又要保证产品的质量。为了保证整车的安全稳定,必须对轮毂的生产和装配提出更高的要求。
剩余内容已隐藏,请支付后下载全文,论文总字数:26303字
该课题毕业论文、开题报告、外文翻译、程序设计、图纸设计等资料可联系客服协助查找;