Document image triangular diagram detection method based on LSD line segment detection

Abstract

The invention discloses a document image triangular diagram detection method based on LSD line segment detection. The method comprises the following steps: 1, carrying out Gaussian filtering and noise reduction on a collected document image, and obtaining a low-noise image; 2, performing LSD line segment detection on the low-noise image to obtain an original line segment set; 3, processing the original line segment set, and combining overlapped and wrongly segmented line segments to obtain a combined line segment set; 4, searching triangular diagram line segment groups meeting a preset condition in the combined line segment set, and performing Harris corner detection on intersection points of straight lines where three line segments in each triangular diagram line segment group are located to obtain triangular diagram corner point groups corresponding to each triangular diagram line segment group; and 5, according to the relative position between each line segment and each angular point in the triangular diagram line segment group and the triangular diagram angular point group, judging whether the position where the triangular diagram line segment group is located is a triangular diagram or not. According to the method, the false detection rate is further reduced, and the triangular diagram in the document image is accurately detected.

Description

technical field [0001] The invention relates to the technical field of image detection, in particular to a document image triangle detection method based on LSD line segment detection. Background technique [0002] Document chart positioning is a key step in the process of screening, retrieving and sorting out document content. The chart positioning work based on electronic format documents mainly judges and positions the charts according to the document data structure. However, in reality, a large amount of electronic literature data is stored in the form of images after scanning paper documents, and it is difficult to apply chart positioning based on electronic formats in this situation. At the same time, the traditional chart positioning work is mainly done manually, which is inefficient. In recent years, people have used computer vision-based detection methods to make up for the shortcomings of traditional manual positioning. The core of such methods is to design fast ...

AI Technical Summary

Problems solved by technology

[0008] At present, the defects of line segment detection mainly include: (1) Due to various reasons such as image collection, the introduction of noisy line segments in the image is inevitable, resulting in a high false detection rate of general detection methods, such as a line segment will be recognized as a multi-segment line segment; (2) While some image detection methods obtain high accuracy, due to the complexity of the algorithm, the detection efficiency is low, and there are certain requirements for computer performance

[0009] The current image detection method has low positioning accuracy for the triangular image in the document image, and some image detection methods have high positioning accuracy, but due to the complexity of the algorithm, the detection efficiency is low, and there are certain requirements for computer performance.

Images