Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/90728
Title: | 改良車載式機器視覺檢測儀量測標線反光性能 Improvement of Vehicle-Mounted Machine Vision Detector for Road Marking Retroreflectivity Measurement |
Authors: | 李佩耘 Pei-Yun Li |
Advisor: | 周家蓓 Chia-Pei Chou |
Keyword: | 道路標線,反光性能檢測,車載式反光儀系統儀器,機器視覺系統,影像檢測, road markings,retroreflectivity performance testing,vehicle-mounted retroreflectometer system,machine vision, |
Publication Year : | 2023 |
Degree: | 碩士 |
Abstract: | 良好的標線反光性能在夜間或光照不足的情況下至關重要,得以確保道路用路者清晰識別標線之所在,並遵照相關指引以保障行車安全。為維護標線性能,除改進標線施作材料及劃設機具,也應加強目前的道路標線檢測和驗證機制,以有效追蹤與管理其效果,有助於推動國內相關的檢測與養護標準的實施。
為增進商用手持式標線回歸反光性能儀器檢測之效率與人員安全性,本研究針對前期學術研究領域相關技術之初步發展,調整原系統中拍攝鏡頭之可視角度,並新增一盞車外額外光源及將整體系統移至檢測車輛前方。在維持原有之各項幾何要求角度下,將原本30公尺外之檢測目標標線前移至距離檢測車前 14 到 16 公尺範圍內,嘗試解決系統易受環境光源干擾而難以檢測之憾。具體目標包括改進拍攝之影像識別效果,改善檢測環境之光源對影像亮度的影響,最後評估標線影像與手持式回歸反光儀量測之關聯。 本文之現地實驗利用額外光源於夜間乾燥天候下,同一檢測路段上分別以開啟與關閉方式進行兩趟次的數據收集,獲得給定距離之兩影像圖像亮度差值除以外加光源於所提供之照度,以計算回歸反射輝度係數,然而,研究發現辨識的精確度受辨識閾值選擇的影響甚鉅,每次更換檢測道路後環境光照度有變化時,閾值和辨識參數必須因地制宜地調整,因此,單一參數難以適用於所有情況,需要透過人工選取的方式進行調整。依據研究結果顯示,後續發展仍有甚大的改進空間,建議後續研究能夠蒐集更多實驗數據,以更全面的了解這兩者之間的關聯,從而進行更深入的分析以求找到更精確的關係。 Ensuring good pavement marking retroreflectivity is crucial for nighttime or low-light conditions, ensuring clear visibility of the markings for road users to follow guidance and ensure driving safety. To maintain marking performance, enhancing in-service road marking inspection mechanisms is essential to effectively monitor and manage their effectiveness. This effort will contribute to advancing the implementation of inspection and maintenance standards related to road markings. To enhance the efficiency and operation safety of the commercial handheld pavement marking retroreflectivity measurement instrument, this study focused on the improvement of the previous relevant technologies in this research field. The modifications included modifying the camera's viewing angle, adding an external light source to the vehicle, and relocating the entire system to the front of the inspection vehicle. While maintaining the original geometric requirements, the detection target that was initially 30 meters away from the system was moved to a range of 14 to 16 meters in front of the inspection vehicle, aiming to address the system's susceptibility to environmental light interference. The specific goals encompassed improving image recognition accuracy, mitigating the impact of ambient light on image brightness, and ultimately evaluating the correlation between pavement marking images and handheld retroreflectivity measurement results. The field experiment in this study involved the use of an additional light source during nighttime dry weather conditions. Two data collection runs were conducted on the same test section, one with the additional light source turned on and the other with it turned off. The obtained difference in image brightness between the two images at a given distance was divided by the illuminance provided by the added light source to calculate the retroreflectance coefficient. However, the study revealed that the accuracy of recognition was heavily influenced by the selection of the recognition threshold. When the ambient light conditions change, the threshold and recognition parameters had to be adjusted accordingly. Therefore, a single parameter was inadequate for all situations, and adjustments needed to be made through manual selection. Based on the research findings, there is still considerable room for improvement in the subsequent development. It is suggested that future research collect more experimental data to gain a more comprehensive understanding of the relationship between these factors. This would enable a deeper analysis to uncover a more precise relationship. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/90728 |
DOI: | 10.6342/NTU202303625 |
Fulltext Rights: | 未授權 |
Appears in Collections: | 土木工程學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-111-2.pdf Restricted Access | 4.28 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.