OCR Layout Analysis Method Based on Gini Impurity

Abstract

The OCR layout analysis method based on Gini impurity of the present invention includes: obtaining an initial OCR identification result including position information and confidence score; dividing all identification results into two categories of high confidence and low confidence according to the confidence score; Starting from the left side of the image, set multiple vertical separation lines at certain intervals. Similarly, starting from the upper side of the image, set multiple horizontal separation lines at certain intervals; use the vertical separation lines to divide the recognition result into two parts: left and right. , the horizontal dividing line divides the recognition result into upper and lower parts; for each dividing line, calculate the Gini impurity of the left and right or the upper and lower parts respectively, and then calculate the total Gini impurity according to the Gini impurity of the upper and lower / left and right; take all the divisions The line with the smallest Gini impurity in the line is used as the final dividing line; the layout of the image is judged by the position and direction of the final dividing line, and the text recognition results on the side with less high-confidence texts are filtered out; The fields are matched to get the final recognition result.

Description

【Technical field】 [0001] The invention mainly relates to the technical field of image processing and OCR, and in particular to an OCR (Optical Character Recognition) layout analysis method based on Gini impurity. 【Background technique】 [0002] OCR is one of the classic tasks in image processing and is widely used in various industries. At present, OCR recognition can achieve high accuracy under the blessing of deep learning algorithm, but the deep learning recognition algorithm itself only outputs all text recognition results in the image, but cannot judge the typesetting information of the image text, let alone the recognition results. Matching with meaningful fields greatly limits the application of OCR in complex scenarios. [Content of the invention] [0003] The purpose of the present invention is to overcome the deficiencies of the prior art, and to provide a layout analysis method for OCR recognition results. Matches provide valid information. In order to realize...

AI Technical Summary

Problems solved by technology

The current OCR recognition can achieve high accuracy with the blessing of deep learning algorithms, but the deep learning recognition algorithm itself only outputs all the text recognition results in the image, and cannot judge the typesetting information of the image text, let alone the recognition results Matching with meaningful fields greatly limits the application of OCR in complex scenarios

Images