基于改进EAST算法的地质图像文本检测

Text detection of geological image based on improved EAST algorithm

针对现有文本检测算法对地质图像中的文本符号检测准确率不够高、不同比例尺的图像检测差异大、文本尺度多样化等问题,提出了一种基于改进EAST算法的地质图像文本检测方法。运用focal-loss函数优化模型损失函数,采用难、易检测文本目标权重动态调整策略来改进文本目标漏检的问题。同时采用多尺度对象分割算法进行图像背景和图像文本的分割,并按比例尺切割训练样本。实验结果表明,对多比例尺栅格地质图像进行文本检测,采用改进后EAST算法的准确率相对改进前提高了5.1%,可更加快速、准确地检索到相关地质图像文字。

In view of the problems that the existing text detection algorithms in geological image detection accuracy are not high enough,the detection difference of different scale images is large,and the text scale is diversified,a geological image text detection method is proposed based on improved East algorithm.The cross entropy strategy is used to improve the loss function to solve the calculation problem of the intersection area of the text rectangle box.The multi-scale method is used to segment the image background and image text.At the same time,the training samples are cut according to the scale.The experimental results show that the accuracy of text detection in geological images is 5.1%higher than that of the original algorithm,and the relevant geological images can be retrieved more quickly and accurately through text symbols.