利用热夜视分割感兴趣区域的调整改进行人检测

2020 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA) Pub Date : 2020-09-23 DOI:10.23919/spa50552.2020.9241295

Karol Piniarski, P. Pawlowski, A. Dabrowski

{"title":"利用热夜视分割感兴趣区域的调整改进行人检测","authors":"Karol Piniarski, P. Pawlowski, A. Dabrowski","doi":"10.23919/spa50552.2020.9241295","DOIUrl":null,"url":null,"abstract":"In this work we present an analysis of the region of interest (ROI) generation from the source thermal night vision images through the double thresholding segmentation technique as a part of pedestrian detection procedure. In some cases, pedestrians do not fit into the generated ROIs. To solve the problem we propose to adjust (slightly enlarge) the segmented ROI. Through this, it is possible to reduce miss rate for the aggregated channel feature (ACF) classifier from 29.1% to 24.8% and for the deep convolutional neural network (CNN) classifier from 24.0% to 22.4%, with negligible impact on the processing time.","PeriodicalId":157578,"journal":{"name":"2020 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2020-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Improved pedestrian detection by adjustment of segmented ROI in thermal night vision\",\"authors\":\"Karol Piniarski, P. Pawlowski, A. Dabrowski\",\"doi\":\"10.23919/spa50552.2020.9241295\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In this work we present an analysis of the region of interest (ROI) generation from the source thermal night vision images through the double thresholding segmentation technique as a part of pedestrian detection procedure. In some cases, pedestrians do not fit into the generated ROIs. To solve the problem we propose to adjust (slightly enlarge) the segmented ROI. Through this, it is possible to reduce miss rate for the aggregated channel feature (ACF) classifier from 29.1% to 24.8% and for the deep convolutional neural network (CNN) classifier from 24.0% to 22.4%, with negligible impact on the processing time.\",\"PeriodicalId\":157578,\"journal\":{\"name\":\"2020 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA)\",\"volume\":\"12 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2020-09-23\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2020 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.23919/spa50552.2020.9241295\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.23919/spa50552.2020.9241295","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

摘要

在这项工作中，我们提出了通过双阈值分割技术从源热夜视图像中产生感兴趣区域(ROI)的分析，作为行人检测过程的一部分。在某些情况下，行人不适合生成的roi。为了解决这个问题，我们建议调整(稍微扩大)分割的ROI。通过这种方法，可以将聚合通道特征(ACF)分类器的缺失率从29.1%降低到24.8%，将深度卷积神经网络(CNN)分类器的缺失率从24.0%降低到22.4%，而对处理时间的影响可以忽略不计。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文本刊更多论文

Improved pedestrian detection by adjustment of segmented ROI in thermal night vision

In this work we present an analysis of the region of interest (ROI) generation from the source thermal night vision images through the double thresholding segmentation technique as a part of pedestrian detection procedure. In some cases, pedestrians do not fit into the generated ROIs. To solve the problem we propose to adjust (slightly enlarge) the segmented ROI. Through this, it is possible to reduce miss rate for the aggregated channel feature (ACF) classifier from 29.1% to 24.8% and for the deep convolutional neural network (CNN) classifier from 24.0% to 22.4%, with negligible impact on the processing time.

求助全文

通过发布文献求助，成功后即可免费获取论文全文。去求助

来源期刊

2020 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA)

自引率

0.00%

发文量