SIA OpenIR  > 光电信息技术研究室
基于机器视觉的枪弹侧表面缺陷检测技术研究
其他题名Bullet side surface detection technology based on machine vision
党领茹1,2
导师朱丹
分类号TP242.62
关键词机器视觉 缺陷检测 特征提取
索取号TP242.62/D23/2014
页数60页
学位专业计算机应用技术
学位名称硕士
2014-05-28
学位授予单位中国科学院沈阳自动化研究所
学位授予地点沈阳
作者部门光电信息技术研究室
摘要近些年,随着科技和生产水平的不断发展,机器视觉技术以其智能高效和非接触式等优点在生产检测中得到了广泛的应用。本文在枪弹表面缺陷检测的应用背景下,开展了基于机器视觉的枪弹侧表面缺陷检测技术研究。 论文内容主要包括:枪弹侧表面缺陷自动检测系统的总体设计,基于HSV颜色模型的单个枪弹分割算法,枪弹侧表面缺陷检测算法以及弹体缺陷分类方法四部分内容。 针对图像中包含多个枪弹目标的情况,本文利用HSV颜色模型中的H分量实现枪弹的分割,并根据枪弹结构空间信息实现枪弹目标的准确定位。 针对枪弹侧表面弹体缺陷和弹尖缺陷各自的特点,分别提出了弹体缺陷检测算法和弹尖缺陷检测算法。弹体缺陷检测算法中,采用了基于光照-反射模型的光照均匀化算法,有效的去除了光照不均的影响;针对亮缺陷和暗缺陷分别设定阈值进行分割;提出可信区域的概念,用来提取有效缺陷目标;利用形态学和区域标记相结合的方法去除小面积结构,达到去除噪声的目的。弹尖检测缺陷检测算法中,利用尖部轮廓的特殊性,采用方向投影的方法获取尖部两个方向的轮廓曲线,通过判断曲线中是否存在极小值点来判定尖部缺陷。 针对枪弹中存在多个缺陷目标的问题,首先将缺陷进行定位,提取出单个缺陷目标子图像。然后,根据枪弹凹坑、裂痕(划痕)、露钢和污渍四类缺陷的特征,选取缺陷图像的几何和灰度特征参数,构造二叉树分类器实现缺陷分类。 为验证本文所提方法的可靠性,在matlab软件环境下,对现有样本进行实验验证。实验结果表明,本文方法能够有效的检测出枪弹侧表面的各种缺陷类型,具有较好的稳定性和鲁棒性。本文的研究成果已经完成硬件实现,并应用在缺陷检测系统当中,检测结果能够满足现阶段的检测需求。
其他摘要In recent years, with the continuous development of technology and production levels, machine vision technology have been widely used in production with its advantages of intelligent, efficient and non-contact detection. In this paper, under the background of the bullet surface defect detection based on machine vision, we have carried out technology research of bullets side surface defect detection based on machine vision The paper mainly includes the following four parts, respectively the overall design of bullet side surface defects automatic detection system, the single bullet extraction algorithm based on HSV color space, the bullet side surface defect detection algorithm and the bullet projectile defect classification method. For the situation of multiple bullets goal in image, the paper implemented the bullet segmentation using H component of HSV color space and realized accurate positioning of bullet target based on the bullet structure space information. For the characteristics of bullet side surface and tip respectively, this paper represented the defects detection algorithm of bullet side surface and the defects detection algorithm of bullet tip. In the defects detection algorithm of bullet side surface, it was used that the light calibration algorithm based on light reflection model to eliminate the influence of uneven illumination. Different threshold was used to segment the light defects and dark defects respectively. The concept of trusted region was improved to extract defect targets effectively. The purpose of removing noise was achieved by using the method of combination of morphology and regional mark. In the defects detection algorithm of bullet tip, for the particularity of contour of bullet tip, direction projection method was adopted to acquire the contour curves, then the defects of bullet tip was determined by judging whether there is a minimum point in curve. For the situation of multiple defect goals in single bullet image, firstly it was realized that extracting a single defect target sub image by locating a defect. According to the characteristics of dents, cracks (scratches), exposed steel and stain, a binary tree classifier was constructed to achieve defect classification with the gray and geometric characteristic parameters of defects In order to verify the reliability of the proposed method, the existing samples were used to experiment in matlab software environment. The experimental results show that this method can effectively detect various types of bullets side surface and is stability and robustness. The research content of this article is implemented by hardware and applied to the defect detection system and the detection results can satisfy the detection requirements of the present stage.
语种中文
产权排序1
文献类型学位论文
条目标识符http://ir.sia.cn/handle/173321/14793
专题光电信息技术研究室
作者单位1.中国科学院沈阳自动化研究所
2.中国科学院大学
推荐引用方式
GB/T 7714
党领茹. 基于机器视觉的枪弹侧表面缺陷检测技术研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2014.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
基于机器视觉的枪弹侧表面缺陷检测技术研究(1556KB) 开放获取CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[党领茹]的文章
百度学术
百度学术中相似的文章
[党领茹]的文章
必应学术
必应学术中相似的文章
[党领茹]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。