Segmentation-based multi-scale feature pyramid text detection method

Abstract

The invention discloses a segmentation-based multi-scale feature pyramid text detection method. The method comprises the steps of obtaining data; constructing a pyramid feature extraction model, and extracting features from the acquired data; sampling input data to obtain input images with different scales, respectively inputting the input images into a pyramid feature extraction model, extractingtext features, fusing and processing the text features of the input images with different scales through a multi-scale detection network to obtain a feature map, and predicting the feature map; and processing a prediction result to obtain a contour boundary line of the text region. The method has high robustness, can be directly applied to text detection of any shape in a natural scene, and achieves high accuracy, recall rate and F value.

Description

technical field [0001] The invention belongs to the field of image text analysis, in particular to a segmentation-based multi-scale feature pyramid text detection method. Background technique [0002] With the development of computer vision technology, the application of image understanding technology is becoming more and more extensive. As an important part of images, text contains rich semantic information and is the key to image understanding. Accurate text detection is the first step to extract key information from images. Text detection in natural scene images faces many challenges due to the diversity of backgrounds, uncertainties in size and direction: (1) the diversity of text formats, and the diversity of text line arrangements; (2) the diversity of text directions ( 3) Diversity of text size and dimension (4) Diversity of text background. [0003] There are currently two main methods for text box detection using deep learning: [0004] (1) Using the anchor-based...

AI Technical Summary

Problems solved by technology

[0006] Among them, the anchor-based method is different from the fixed aspect ratio of general objects due to the variety of scene text sizes, which makes the network insensitive to the size of the text and has low accuracy; in addition, most of the existing anchor-based text detection is Based on quadrilaterals or rotated rectangles, text of arbitrary shape cannot be positioned well

The method based on pixel segmentation is easily limited by the receptive field, which is not good for small target detection, and for some relatively close text instances, it cannot distinguish the boundary well

Images