Chinese positioning, segmenting and identifying method in natural scene image

Abstract

The invention provides a Chinese positioning, segmenting and identifying method in a natural scene image. Through a FASText model, an original picture is subjected to preliminary character positioning, a candidate character area is extracted, the candidate character area is pre-segmented, the separate word part of the pre-segmented character area is identified, and a field part can be subjected to further separate-word segmentation and identification. By use of the method, through the accurate extraction of character stroke characteristics and the powerful character identification ability of a deep residual neural network, a path tree method is combined to simply and effectively realize a purpose of Chinese positioning and identification, and the method can be applied to various natural scenes and does not need supervised training.

Description

Technical field

[0001] The invention belongs to the technical field of image processing, and specifically relates to a method for Chinese positioning, segmentation and recognition in natural scene images. Background technique

[0002] Text recognition in natural scenes is a very important visual detection target. The text in the image contains a lot of useful information, which is essential for the understanding and acquisition of visual content. There are many related text recognition applications, including road signs, license plates, bills, and so on.

[0003] Generally speaking, traditional OCR technology is affected by the complex background of natural scenes, and it is difficult to correctly complete related tasks. On the whole, this type of task can be divided into two stages, text localization and recognition. Text positioning is the precise positioning of the text position in the image, which is mainly based on extracting relevant text features, such as MSERs, to disting...

Images