結合AI平台與IoT識別河川水位與都市淹水

近年來全球受極端氣候變遷影響，都市街道或道路發生間歇性暴雨，容易導致市區道路造成嚴重的積淹水事件。本文藉由OpenCV軟體影像辨識、人工智慧之深度學習及物聯網（IoT）等技術，進行自動化河川水位、淹水辨識及淹水深度計算。利用python程式語言，建置AI平台自動化展示系統，經實驗室模擬、歷史颱洪事件及現場影像分析驗證結果顯示，於日、夜及颱洪期間均適用水位計算，亦可提供無水位計處之即時估算水位；另利用人工智慧深度學習之卷積神經網路（CNN）inception v3遷移學習的訓練與預測，準確識別淹水影像，即時計算淹水深度，可彌補人工無法長時間判識淹水與否的缺點。本文成果可提供災前整備、災中應變之影像圖資，以達自動化即時河川水位及淹水深偵測判識之目標。

關鍵字

人工智慧；卷積神經網路；影像辨識；物聯網

並列摘要

In recent years, the world has been affected by extreme climate changes, and intermittent rainstorms have occurred on urban streets or roads, which can easily cause serious flooding events on urban roads. With OpenCV software image recognition, deep learning of artificial intelligence and internet of things (IoT) technologies, automatic river water level, flood recognition and flood depth calculations are performed. Using python programming language, self-built AI platform automatic display system, the laboratory simulation, historical Taiwan flood events and on-site image analysis verification results show that the water level calculation is applicable during the day, night and during the flood period, can also be provided estimation of real-time water level without water level gauge. In addition, use the convolutional neural network (CNN) inception v3 migration learning training and prediction of artificial intelligence deep learning to accurately identify whether the image is flooded or not, and calculate the flooding depth in real time, which can compensate for the artificial inability to judge for a long time. Know the shortcomings of flooding or not. This results can provide river stage and inundation depth data for pre-disaster preparation, and disaster response, achieving the purpose of automatic real-time detection and identification.

並列關鍵字

Artificial intelligence ； Convolutional neural networks (CNN) ； Image identification ； Internet of things (IoT)

國際替代計量

結合AI平台與IoT識別河川水位與都市淹水

全文下載

主題瀏覽