在數位典藏的推廣應用中，對於中文古籍全文資料庫的建立，古籍文字切割與辨識是重要環節。光學字元辨識(Optical Character Recognition, OCR)，需要切割出每一個文字的座標再加以辨識，切割文字錯誤時，辨識的結果也就不可能正確。因此文本的文字切割對於正確的文字識別有著重大的影響和價值，本研究我們設計了一個中文古籍文本的文字切割系統。我們設計的系統輸入影像為彩色掃瞄圖檔，我們使用CRAFT文字偵測模型對輸入影像進行文字偵測，找出文字像素範圍，加以切割出文字區塊圖片。;In the promotion and application of digital archives, for the establishment of the full-text database of Chinese ancient books, the text cutting and identification of ancient books is an important link. Optical Character Recognition (OCR) requires cutting out the coordinates of each character and then recognizing it. When the character is cut wrong, the recognition result cannot be correct. Therefore, the text cutting of the text has a significant impact and value on the correct text recognition. In this research, we have designed a text cutting system for Chinese ancient texts. The input image of the system we designed is a color scan image file. We use the CRAFT text detection model to detect the text of the input image, find the text pixel range and then cut out the text block image.
