本篇論文利用特殊螢光蛋白與待測食品混合後的螢光表現量，來判定食品是否具有毒性。然而，傳統固定式實驗室所使用的分析儀器，價格昂貴且不具備可行動性，過程費時，無法即時產生檢測結果，故本論文結合IoTtalk物聯網平台及具備光學相機之智慧型手機，利用行動裝置便於攜帶且容易使用的特性，可即時擷取螢光反應結果，再藉由物聯網平台與後端毒性化合物篩檢伺服器判斷並回覆初步檢測結果。為了讓手機也具備可參考的檢測準確率，本文提出一二階段校正學習及光學影像辨識檢測演算法。第一階段將取樣校正照片並計算校正曲線與學習，第二階段則利用前階段的學習結果對待測物的螢光反應結果進行辨識與檢測。無論第一或第二階段，我們的方法均會對照片使用灰卡平衡、白平衡或灰度世界演算法進行校正，以修正照片的顏色偏差，然後對待測區域進行數值加權算法。本文所提出結合物聯網平台與行動裝置之毒性化合物初步篩檢系統，不僅提供操作人員與使用者行動篩檢平台，同時保證了檢測所需的即時性與專門性。

This paper uses the fluorescence expression of special fluorescent protein (which is mixed with the food sample) to determine whether the food is toxic or not. Compared to traditional detection and judgment procedures, they are usually expensive, not portable and quite time consuming. Therefore, this paper combines the IoTtalk (an IoT platform) and the smart phone with optical camera to realize a “portable” analytical system. Exploiting the easy to carry and easy to use characteristics of smartphone, users can capture the fluorescence reaction results in real time, and then send the picture to the IoT platform. Upon the receipt of the picture, the platform will send the picture to the application server, the toxic compound screening server, for testing results and determining whether the food sample is toxic or not. To realize the idea and make the testing result reliable, we propose a two stage learning based optical image recognition detection algorithm for preliminary screening. In the first stage, pictures of different fluorescence concentration samples are captured for calibration and model training. Then, the second stage will exploit the trained model in the first stage to detect and screen the fluorescent reaction results of the testing object. Our method will apply gray card balance, white balance, or gray world algorithm for calibration to correct the color deviation of photos and then execute the numerical weighting algorithm to the testing area of the photo in both the first and second stage. Integrating the Internet of Things (IoT) platform and mobile devices, we propose a preliminary screening system for toxic compounds in this work. The system provides operators and users a mobile screening platform. Thus, inspections can be done at any time and any place. The proposed mobile screening service is easy to operate for users while ensuring the required immediacy and correctness.