Handwritten ancient character detection method

The invention provides a handwritten ancient character detection method, which comprises the steps of performing feature learning and character classification on an input image and corresponding annotation information by using an ATD network based on a CNN, and generating an AT type candidate box according to a classification result; preprocessing an input image through non-local mean filtering byusing an MSER model based on an NMS, extracting a text contour by using an MSER algorithm, generating candidate boxes by using a minimum bounding rectangle, and screening out the most accurate MT candidate boxes by the NMS; and synchronously outputting two different candidate boxes of the same character by the ATD network and the MSER model, and combining the two candidate boxes by a combinationalgorithm to obtain an FT type textbox. Test results on ancient book data sets such as Yi language, Chinese, Latin language, Italian language and the like show that the method provided by the invention has good precision and m