Edge Detection using CNN for Roof Images

Aneeqa Ahmed, Y. Byun
{"title":"Edge Detection using CNN for Roof Images","authors":"Aneeqa Ahmed, Y. Byun","doi":"10.1145/3314527.3314544","DOIUrl":null,"url":null,"abstract":"In image processing detection of edges is a very significant phenomenon as it finds its utility in multiple applications from different areas of life such as security, video surveillance, detection of various diseases, text detection, vehicle detection and many more. Image processing based techniques for edge detection usually requires preprocessing of the images in order to reduce the quantity of noise before edge detection. Even after the detection of edges, most of the times these techniques have to involve various post processing steps in order to generate the fine edges to be used in concerned applications like tumor detection or critical security applications. Image processing based algorithms for edge detection are conventionally based on gradient filters and generally they are not reliable in the case of variable illumination conditions or noise deviations. As soon as there is any variation in the illumination conditions of images or change in the percentage of noise element in images, the efficiency of the traditional edge detection algorithms is affected badly because of increased mask size. This increased mask also results in additional computational load. This paper utilizes Convolutional Neural Networks (CNN) for the detection of edges in images. The dataset composes roof images of diverse types and sizes. Employing CNN for edge detection over conventional methods is more reliable and it is speedier and less complex. Also, CNN is very efficient as it can take compute all features automatically so reduces the burden of manual feature extraction. Moreover, feature detection using CNN is accurate and more precise than manual feature compilation methods. To make use of automatic feature extraction of CNN, this technique employs Visual Geometry Group (VGG) CNN network. Further, to compute the edge map, Roberts edge operator is applied on the automatically computed features. This CNN based edge detection technique results into the generation of very fine edges without the application of any post processing step. Also, the results show that this technique works equally good in the presence of noise and requires no pre-processing algorithms.","PeriodicalId":370433,"journal":{"name":"Proceedings of the 2019 Asia Pacific Information Technology Conference","volume":"4 3 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2019-01-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the 2019 Asia Pacific Information Technology Conference","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/3314527.3314544","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1

Abstract

In image processing detection of edges is a very significant phenomenon as it finds its utility in multiple applications from different areas of life such as security, video surveillance, detection of various diseases, text detection, vehicle detection and many more. Image processing based techniques for edge detection usually requires preprocessing of the images in order to reduce the quantity of noise before edge detection. Even after the detection of edges, most of the times these techniques have to involve various post processing steps in order to generate the fine edges to be used in concerned applications like tumor detection or critical security applications. Image processing based algorithms for edge detection are conventionally based on gradient filters and generally they are not reliable in the case of variable illumination conditions or noise deviations. As soon as there is any variation in the illumination conditions of images or change in the percentage of noise element in images, the efficiency of the traditional edge detection algorithms is affected badly because of increased mask size. This increased mask also results in additional computational load. This paper utilizes Convolutional Neural Networks (CNN) for the detection of edges in images. The dataset composes roof images of diverse types and sizes. Employing CNN for edge detection over conventional methods is more reliable and it is speedier and less complex. Also, CNN is very efficient as it can take compute all features automatically so reduces the burden of manual feature extraction. Moreover, feature detection using CNN is accurate and more precise than manual feature compilation methods. To make use of automatic feature extraction of CNN, this technique employs Visual Geometry Group (VGG) CNN network. Further, to compute the edge map, Roberts edge operator is applied on the automatically computed features. This CNN based edge detection technique results into the generation of very fine edges without the application of any post processing step. Also, the results show that this technique works equally good in the presence of noise and requires no pre-processing algorithms.
利用CNN对屋顶图像进行边缘检测
在图像处理中,边缘检测是一个非常重要的现象,因为它在生活的各个领域都有应用,如安防、视频监控、各种疾病的检测、文本检测、车辆检测等等。基于图像处理的边缘检测技术通常需要在边缘检测之前对图像进行预处理,以减少噪声的数量。即使在边缘检测之后,大多数情况下,这些技术必须涉及各种后处理步骤,以生成用于肿瘤检测或关键安全应用等相关应用的精细边缘。基于图像处理的边缘检测算法通常是基于梯度滤波器的,通常它们在可变照明条件或噪声偏差的情况下不可靠。一旦图像的光照条件发生变化或图像中噪声元素的百分比发生变化,由于掩模尺寸的增大,传统边缘检测算法的效率会受到严重影响。这种增加的掩码也会导致额外的计算负载。本文利用卷积神经网络(CNN)对图像进行边缘检测。该数据集由不同类型和大小的屋顶图像组成。采用CNN进行边缘检测比传统方法更可靠、更快、更简单。此外,CNN非常高效,因为它可以自动计算所有特征,从而减少了手动特征提取的负担。此外,使用CNN进行特征检测比人工特征编译方法更准确、更精确。为了利用CNN的自动特征提取,该技术采用了VGG (Visual Geometry Group) CNN网络。为了计算边缘映射,在自动计算的特征上应用Roberts边缘算子。这种基于CNN的边缘检测技术在没有任何后处理步骤的情况下生成了非常精细的边缘。此外,结果表明,该技术在存在噪声的情况下同样有效,并且不需要预处理算法。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信