Character Recognition for Automatic Draught Survey

摘要

Draught survey is one of the widely used techniques to measure the amount of bulk cargo in freighter ships. To support the operator responsible for the draught reading process, as well as, to ensure the process impartiality and accuracy, this paper presents the development of a method, based on computer vision techniques, able to track, segment and recognize the draught marks. The proposed method was developed using techniques for feature extraction, edge detection and optical character recognition. It was tested using images from different ships taken at Tubarão Port in Vitoria - ES and at Madeira`s Maritime Terminal in São Luis - MA. Those images were taken from different positions at distinct hours of the day. For real-world images, which has directly aligned draught marks, the proposed method has accomplished an accuracy of 94%, what shows that the algorithm works satisfactorily under these conditions. Another aspect that raises even more the importance of this paper is the lack of others that aim to solve the same problem. Automatization of draught survey can bring great benefits to the bulk cargo loading process, which is of utmost importance to the global market.