Correction Method of String Direction in Drawings Based on Joint Features

Abstract

The invention discloses a method for correcting the direction of character strings in drawings based on joint features. Randomly select two different points to calculate the direction angle of each character; calculate the side length ratio of the character according to the long side and short side of the smallest circumscribed rectangle of the character, and mark the characters whose side length ratio is greater than or equal to the preset threshold is the first type of character; based on the character Hu invariant moment, extract the moment feature of the first type of character, and according to the moment feature, use the trained SVM classifier to classify each character; update its direction angle according to the classification result; according to When the number of characters in the string is the same as the preset number, the main direction rotation angle of the string is obtained by using the K-means clustering algorithm or the arithmetic mean, and then the main direction rotation angle of the string is used to correct the main direction of the string.

Description

technical field [0001] The invention relates to image processing technology, in particular to a method for correcting the direction of character strings in drawings based on joint features. Background technique [0002] Common drawings include engineering drawings and mechanical drawings. Engineering drawings are an important basis for engineering design, construction and acceptance. Mechanical drawings are an important basis for the design, processing and quality inspection of mechanical parts. The character annotation in each type of drawing is a key data in the drawing. The foreground of the image in the drawing is complex, but the background is relatively simple, and there is no rich color information; the annotation strings are unevenly distributed, with different directions, and are interspersed in various positions of the image. Automatic extraction and recognition of character annotation strings can help people quickly identify, understand, and manage drawings, as...

AI Technical Summary

Problems solved by technology

The operation of character detection and string segmentation is likely to cause distortion of the shape of the character and distort the original direction of the character

[0005] (3) Character correction of label strings. In order to compactly arrange the layout in the drawing and facilitate the user's understanding, the character direction of the label string in the drawing is very flexible. Strings in all directions may exist, which allows the computer to automatically extract and recognize the marked characters string causing difficulty

[0009] The disadvantage of the tilt correction method based on the straight line detection of text lines is that the effect is not good for the more complex arrangements in engineering drawings that do not form text lines; the disadvantage of the tilt correction method based on corner points and projections is that the character arrangement in engineering drawings There may be multiple lines of misalignment, which cannot maintain good robustness; the disadvantage of the existing distortion correction method based on deep learning is that for engineering drawings, the characters have free direction rotation instead of distortion, so this method is not feasible. High, and deep learning training takes a long time, requiring a lot of resource overhead, not economical enough

Images