基于PSENet的人民币文本检测方法

RMB Text Detection Based on PSENet Algorithm

21世纪以来,人民币文本检测得到了快速发展,然而人民币文本检测仍存在很多问题:一方面,人民币存在破损、折角、文字模糊等问题,导致难以检测人民币图像文本区域;另一方面,拍摄采集的人民币图像易受到光照强度、图像变形、拍摄角度等多方面的影响,导致文本区域定位的精准度不高。针对以上问题,本文提出了将PSENet应用于人民币文本检测的方法,PSENet可以对任意形状的人民币文本进行定位,提高文本区域定位的精准度,PSENet可以将相邻人民币文本分离,提高人民币文本检测的准确率。经实验结果表明,基于PSENet的人民币文本检测方法的准确率为98%,具有较高的参考价值,因此本文将PSENet应用于人民币文本检测。

Since the 21st century,RMB text detection has developed rapidly.However,there are still many problems in RMB text detection:on the one hand,there are some problems such as RMB damage,corner,text blur,etc.,which make it difficult to detect the text area of RMB image;on the other hand,the captured RMB image is easily affected by illumination intensity,image deformation,shooting angle and so on The accuracy of text region location is not high.In view of the above problems,this paper proposes a method to apply psenet to RMB text detection.Psenet can locate any shape of RMB text and improve the accuracy of text region positioning.Psenet can separate adjacent RMB text and improve the accuracy of RMB text detection.The experimental results show that the accuracy of RMB text detection method based on psenet is 98%,which has high reference value.Therefore,this paper applies psenet to RMB text detection.