利用機器學習於邊坡光學與紅外線熱成像融合之長期穩定性監測評估;The evaluation of slope monitoring using optical and thermal images fusion through machine learning

NCUIR > Engineering College > Graduate Institute of Civil Engineering > Electronic Thesis & Dissertation > Item 987654321/91903

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/91903

Title:	利用機器學習於邊坡光學與紅外線熱成像融合之長期穩定性監測評估;The evaluation of slope monitoring using optical and thermal images fusion through machine learning
Authors:	曾得瑋;Tseng, Te-Wei
Contributors:	土木工程學系
Keywords:	光學成像;紅外線熱成像;機器學習;變化檢測;邊坡災害防治;Optical imaging;Infrared thermal imaging;Machine learning;Change detection;Slope monitoring
Date:	2023-07-28
Issue Date:	2024-09-19 14:42:10 (UTC+8)
Publisher:	國立中央大學
Abstract:	山坡地的變化和移動量始終為其是否會坍塌的重要因素，以安全性高、安裝成本低、機動性高的數位攝影測量方式，長時間地對山坡地進行拍攝及觀測，可以有效地確保山坡地的穩定性。本研究旨在將光學成像技術、紅外線熱成像技術與機器學習方法結合，並且用於雙時段圖像的變化檢測上。利用微型單板電腦—樹梅派結合兩鏡頭模組，通過在現場同一位置但不同時間拍攝的圖像，以機器學習的方式，讓電腦預測雙時段光學圖像組中的變化位置;以數值處理的方式，讓電腦計算雙時段紅外線熱圖像組中的溫度變化。本研究先經室內模型實驗，使用自行收集的圖像組，訓練機器學習模型，以此對變化檢測模型進行可行性分析和檢驗，後再將監測系統移至室外，在現有坡地旁架設測站，根據監測結果顯示，本研究之方法能夠有效地在室外環境下進行長時間地監測作業，在國立中央大學停車場變化檢測的 mean F1-score 和 mean IoU 分別能達到 0.9359 和 0.8850，而在新北市瑞芳區南雅里的成效良好。 ;The change and displacements of slopes has always been an important factor in determining whether they will collapse. Digital photogrammetric methods, which are characterized by high safety, low installation cost, and high mobility, can effectively ensure the stability of slopes by continuously capturing and observing them over a long period of time. This study aims to combine optical imaging technology, infrared thermography, and machine learning methods for change detection in dual-temporal images. By utilizing a micro single board computer, specifically the Raspberry Pi, in combination with two camera modules, images are captured at the same location but at different times. Through machine learning, the computer can predict the locations of changes in the bi-temporal optical image set. Numerical processing is used to calculate temperature changes in the bi- temporal infrared thermographic image set. Initially, indoor model experiments were conducted using a self-collected set of images to train the machine learning model and analyze the feasibility and effectiveness of the change detection model. Subsequently, the monitoring system was deployed outdoors, setting up a monitoring station near existing slopes. The monitoring results demonstrated that the proposed method in this study can effectively perform long-term monitoring operations in outdoor environments. The change detection score of mean F1-score and mean IoU in the parking lot of National Central University can reach 0.9359 and 0.8850 respectively, showcasing good performance in the Nanya Village, Ruifang District, New Taipei City.
Appears in Collections:	[Graduate Institute of Civil Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	36	View/Open

社群 sharing

Loading...