Test report structured recognition method based on deep learning and automatic error correction

The invention relates to the technical field of medical laboratory sheet identification, in particular to a test report structured identification method based on deep learning and automatic error correction, which improves the scene characteristics of medical laboratory sheet detection and identification based on a differentiable binary network. The method comprises the following steps: performing text detection and recognition on a test report image by using a differential binary network of cascade sparse query to obtain a structured recognition result of a test report, and in addition, aiming at the problems of project misalignment, single-project information separation, multi-project information adhesion and the like existing in the structured result output by an OCR module, through a BK tree and an AC automaton, obtaining the structured recognition result of the test report. And performing automatic error correction on the item information in the organized structured table, finally combining the identifie