[iOS]Chinese Mix English recognization

  • 3.9K Views
  • Last Post 14 September 2013
BabyKen posted this 03 September 2013

I tried the OCR mobile engine. So far the result is not as good as I expected. I have attached an ID card example. Simply I would like to read the

Chinese name "李智能"

English Name "Lee, Chi Nan" (or "Lee Chi Nan" if it could, probably using regular expression)

The Chinese name code ("2621 2535 5174") in this case

Date of birth : "01-01-1968"

Sex : "女"

Sex : "F"

the HKID : C668668(E)

I tested the Business card features and here is my result

2621 2535 517

Phone: : 01-79

Phone: : 15-09-03

Mobile: : 668668

Address: : 香液永久性居民#份義

Address: : HCING KONG� rERMAI�tINT ID1:NT1TV CARD

Address: : ,+,生g期DBteofBlml

Address: : 01-D卜19.6.S

Address: : ***AZ -_.

Address: : ggg隅Dalgoua

Name: : 樣本

Name: : Lee/ Chi Nan

Company: : 女F

Other Text: : 香液永久性居民#份義

Other Text: : HCING KONG� rERMAI�tINT ID1:NT1TV CARD

Other Text: : 樣本SAMPLE

Other Text: : LEE/ Chi Nan

Other Text: : 2621 2535 517�

Other Text: : ,+,生g期DBteofBlml

Other Text: : 01-D卜19.6.S

Other Text: : ***AZ -_.

Other Text: : ggg隅Dalgoua

Other Text: : ♦◎

Other Text: : 女F

Other Text: : l

Other Text: : 旧

Other Text: : (01-79)

Other Text: : 15-09-03

Other Text: : C668668<E)

Other Text: : 、

Other Text: : J

Here are my questions.

  • How to determine the position / location to detect and match with its index. E.g. How do it know the "address" located in ?

  • Is there any setup for detecting Chinese? Except 女 and 樣本, rest of the Chinese are incorrect.

Please kindly give me an aid on ID card detection. I believe your product should able to detect it without any problems. Thanks in advance.

Attachment is ID card example.

alt text

Attached Files

Anastasia Galimova posted this 14 September 2013

Firstly we recommend to improve the image quality: the height of the hieroglyphs on the attached image is about 11-16 pixels, while at least 30-35 is required.

If the issue still occurs, please share or send to CloudOCRSDK@abbyy.com the image you recognize and we will try to fix it in the future.

As a temporary solution you can try to improve the recognition quality by specifying text coordinates directly (use the processFields method for it).

Close