An Automatic System for Container Code Location Extraction and Recognition

[[abstract]]高雄港貨櫃裝卸量於1999年排名世界第三，自動化貨櫃碼辨識系統對港口貨櫃進出口管理絕對有相當程度的幫助，為使貨櫃管理更自動化，本研究即利用數位影像處理技術來辨識貨櫃碼，以節省人工作業所耗費的時間及人力成本。貨櫃碼共11碼，分別由3碼貨櫃擁有者英文代碼、1碼貨櫃類別英文代碼、6碼數字序號及1碼檢核碼組成，本研究的辨識方法主要可以分為以下四個步驟：第一、貨櫃為彩色塗刷，而多樣性的顏色對貨櫃碼辨識並沒有幫助，故本研究藉由前處理先將貨櫃影像轉成只有黑、白的二值化影像，進而利用邊緣化、八連通等方法對貨櫃碼定位，而後抽取物件。第二、因貨櫃碼印刷有單列、雙列印刷，故本研究透過群組化方式，將位於同一水平的貨櫃碼群集並排序，再透過水平投影、垂直投影切割出貨櫃碼字元影像，並將這些切割出來的字元影像大小予以正規化。第三、在切割出字元影像後，利用樣板比對來辨認號碼，考量貨櫃來自世界各國，字型也不盡相同，故採取多重樣版來因應字型問題，另外本研究亦使用三碼比對法，將待辨識影像及樣板切割為三等份，解決辨認過程的混淆字問題。第四、為提供辨識錯誤的補救，本研究將原始貨櫃碼於資料庫中建檔，得到辨識結果後藉以驗證辨識正確性，同時也提供辨識錯誤的補救。最後為驗證本研究所提出系統之有效性，採用140張貨櫃影像來進行實驗，實驗結果顯示所提之方法確實可快速並能正確地從影像中辨認出貨櫃號碼。[[abstract]]The container traffic of the Port of Kaohsiung was ranked third in the world Automatic container code recognition (ACCR) was quite helpful to manage the container import/export business from the harbor In this study digital image processing technique is used to automatically recognize the container codes characters (CCCs) which consist of 11 characters and can be divided into four English letters six digital numbers and one check digital number The procedure can be divided into the following four steps: First the image preprocessing involves transforming a color image into a gray-level and binarizing image applying edge detection and performing 8-connected component to locate the CCCs’ region Second using vertically and horizontally projection of container code and the segmented character normalization the double- and single-row CCCs can be extracted and ordered from the input images Third this study combines a template-matching and three-section matching method to recognize the extracted characters Final in order to increase the recognition rate this study also try to estimate some mis-recognized characters using the prestored database Moreover 140 container code images have been used to test the proposed method The results show that the proposed method is more effective and accurat