利用紅外線熱成像強化混凝土橋梁表面剝落檢測成效之初步研究

Abstract

橋梁老化產生各種損傷和缺陷，會影響橋梁結構安全，因此，定期巡檢、有效管理、及時維護和補強是確保橋梁使用安全和品質的重要措施。目前世界各國仍普遍採用目視檢測作為橋梁檢測最主要的方式，惟目視檢測仰賴檢測人員的訓練素質、經驗和主觀認定，易產生判斷不一之情形。 國內外多項研究以深度學習技術應用於橋梁檢測，作為偵測裂縫的非接觸式量測方式，惟迄今AI影像辨識技術尚有瓶頸待突破。而過去文獻指出紅外線熱像儀可有效偵測混凝土橋梁表面剝落，本研究旨在探討以較低規格之紅外線熱像儀檢測混凝土橋梁表面剝落及輔助AI影像辨識技術，應用於橋梁檢測是否具有可行性。 經實驗室研究與橋梁實拍結果顯示，本研究所採用之紅外線熱像儀(FLIR E5)，最適當量測距離為1至2米，且可偵測出各式各樣形狀的剝落，但較不適合量測深度較淺及面積較小之缺陷尺寸；裂縫周圍有局部剝落時，有助於提高紅外線熱成像之偵測率，卻會降低AI影像辨識的偵測率；而AI影像辨識對於假缺陷之誤判率遠大於紅外線熱成像，因此利用較低規格之紅外線熱像儀檢測混凝土橋梁表面剝落及輔助AI影像辨識技術，應用於橋梁檢測具有可行性。

Regular inspections, effective management, and timely maintenance are critical issues to ensure bridge safety and quality. Currently, visual inspection remains the predominant method employed worldwide for bridge inspection. However, visual inspection heavily relies on the training, experience, and subjective judgment of inspectors, leading to inconsistent assessments. Several studies have utilized deep learning to detect cracks during bridge inspections. Nevertheless, there is still a significant challenge to overcome AI image recognition. Previous literature has highlighted the effectiveness of thermal imaging cameras in detecting concrete bridge surface spalling. This study aims to investigate the feasibility of employing low-standard infrared thermal imaging cameras to detect concrete bridge surface spalling and support AI image recognition technology in bridge inspections. Based on laboratory research and on-site inspections of bridges, the FLIR E5 infrared thermal imaging camera has demonstrated the recommended measurement distance of 1 to 2 meters. While it may not be suitable for measuring shallow-depth and small-area defects, it excels in effectively detecting various shapes of spalling and supporting AI image recognition. Consequently, employing low-standard infrared thermal imaging cameras for the detection of concrete bridge surface spalling and integration with AI image recognition technology in bridge inspections appears to be a feasible approach.