以連續角度光源及卷積神經網路檢測光影變化箔膜影像

Use Continuous Angle Light Source and Convolutional Neural Network to Identify Optically Variable Device

光影變化箔膜常用於鈔券及安全文件，由於其箔膜上的影像內容及色彩會隨著觀看者角度而產生變化，鑑識人員不易於不同觀看角度時記住每一角度光影變化箔膜所呈現的影像及色彩。因此本研究採用了連續角度光源照射於光影變化箔膜以得到每一角度的影像變化，再經由卷積神經網路訓練和測試，可成功且精準的將不同的光影變化箔膜加以辨識和分類。

Optically Variable Device is often used in banknotes and security documents. Because the image content and color on the OVD will change with the angle of the viewer, it is not easy for forensics personnel to remember the unique image and color changes at each angle of light source. Therefore, this study used a continuous-angle light source to illuminate the OVD to obtain the image changes at each angle, and then through the convolutional neural network training and validation, different OVDs can be successfully and accurately identified and classified.