In our order recording system, in addition to uploading the ID card, it is also necessary to manually fill in the ID card and bank card number, especially the bank card number. Those bump effects and special fonts are really unfriendly to the human eye. There are many old orders in the previous system. The card number was entered incorrectly. Look at the ID card first.
For the identification of ID card content, input OCR technology, there are some paid and open source implementations.
Among the open source implementations, the most famous one is tesseract. The corresponding website is https://github.com/tesseract-ocr/tesseract. In addition to recognition, different language libraries can be added for training to improve the recognition rate. For the usage of tesseract, see http://www.52itstyle.com/thread-4803-1-1.html, which explains it clearly.
In the process of testing, I found that it is not very stable, and the recognition rate is greatly affected by the quality of the picture, especially for some white light or backlight pictures, many of which cannot be recognized. If you understand the knowledge of image processing, you can make some necessary corrections to the picture first, which should greatly increase the recognition rate and achieve the purpose of being usable.
I also searched for the implementation of fees, such as Yunmai, Megvii, etc. Among them, Megvii http://cloud.megvii.com is a paid product, mainly including face, ID card, driver's license and other related recognition applications, among which ID card recognition has free http service and sdk for android and ios, There are many effects, basically 100% recognition, but it only includes the recognition of network pictures, and there are restrictions on local pictures.