Altitude Analysis of Road Segmentation from UAV Images with DeepLab V3+

2022 IEEE 12th International Conference on Control System, Computing and Engineering (ICCSCE) Pub Date : 2022-10-21 DOI:10.1109/ICCSCE54767.2022.9935649

Mat Nizam Mahmud, Muhammad Hiszarul Azim, M Fazmi Hisham, M. K. Osman, A. P. Ismail, F. Ahmad, K. A. Ahmad, A. Ibrahim, Azmir Hasnur Rabiani

{"title":"Altitude Analysis of Road Segmentation from UAV Images with DeepLab V3+","authors":"Mat Nizam Mahmud, Muhammad Hiszarul Azim, M Fazmi Hisham, M. K. Osman, A. P. Ismail, F. Ahmad, K. A. Ahmad, A. Ibrahim, Azmir Hasnur Rabiani","doi":"10.1109/ICCSCE54767.2022.9935649","DOIUrl":null,"url":null,"abstract":"DeepLab V3+ semantic segmentation develops road segmentation from UAV images. First, a camera-equipped UAV captures road images from 3 altitudes in Perlis. The images will be resized and augmented to provide additional road images for deep learning model training. Next, images are manually segmented into road and background using CVAT. The DeepLab V3+ with Resnet-18, Resnet-50, and MobileNet V2 backbone network is utilised to segment the road using Matlab. Finally, the suggested method's performance is compared to all backbone network approaches at 3 various altitudes to determine pixel accuracy (PA), mean intersection over union (mIoU), and meanF1-score (meanF1). The study develops an accurate and robust approach for road segmentation from UAV images that road surveyors may employ for inspection and monitoring. This technique might be implemented to identify road cracks and potholes in the future study.","PeriodicalId":346014,"journal":{"name":"2022 IEEE 12th International Conference on Control System, Computing and Engineering (ICCSCE)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE 12th International Conference on Control System, Computing and Engineering (ICCSCE)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICCSCE54767.2022.9935649","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

DeepLab V3+ semantic segmentation develops road segmentation from UAV images. First, a camera-equipped UAV captures road images from 3 altitudes in Perlis. The images will be resized and augmented to provide additional road images for deep learning model training. Next, images are manually segmented into road and background using CVAT. The DeepLab V3+ with Resnet-18, Resnet-50, and MobileNet V2 backbone network is utilised to segment the road using Matlab. Finally, the suggested method's performance is compared to all backbone network approaches at 3 various altitudes to determine pixel accuracy (PA), mean intersection over union (mIoU), and meanF1-score (meanF1). The study develops an accurate and robust approach for road segmentation from UAV images that road surveyors may employ for inspection and monitoring. This technique might be implemented to identify road cracks and potholes in the future study.

查看原文

微信好友朋友圈 QQ好友复制链接

本刊更多论文

基于DeepLab V3+的无人机图像道路分割高度分析

DeepLab V3+语义分割基于无人机图像进行道路分割。首先，一架装有摄像头的无人机从3个海拔高度拍摄了波斯的道路图像。这些图像将被调整和增强，为深度学习模型训练提供额外的道路图像。接下来，使用CVAT将图像手动分割为道路和背景。采用带Resnet-18、Resnet-50和MobileNet V2骨干网的DeepLab V3+，利用Matlab对道路进行分段。最后，将该方法与所有骨干网方法在3个不同高度的性能进行比较，以确定像素精度(PA)、平均交联(mIoU)和均值f1 -score (meanF1)。该研究开发了一种准确而稳健的方法，用于从无人机图像中分割道路，道路测量师可以使用该方法进行检查和监测。在未来的研究中，这种技术可能被用于识别道路裂缝和坑洼。

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分钟内获得全文去求助

来源期刊

2022 IEEE 12th International Conference on Control System, Computing and Engineering (ICCSCE)

自引率

0.00%

发文量