Business card identifying method combining character identification with image matching

Abstract

The invention provides a business card identifying method combining character identification with image matching, which is applied to a network system constructed by a website server and a client. The business card identifying method at least comprises the following steps of: retrieving a business card image of a request business card by the client and uploading the business card image to the website server to request for identification; extracting character information and image characteristics on the business card image by the website server, then retrieving a prestored business card matched with the character information, carrying out image characteristics matching to screen out candidate business cards, and generating a metric value corresponding to each candidate business card; judging whether a matched result is determined according to the metric value by the website server; if so, feeding back the matched result to the client; if not, sending the business card images of the candidate business cards to a service side authorized by the website server through a network; and judging the matched result by the service slide and then feeding back information to the client, therefore, the business card identifying technology combining character identification with image matching can be realized so as to improve the accuracy rate and the speed of business card identification.

Description

technical field [0001] The invention relates to the technical field of image processing and network application, in particular to a business card recognition method combining character recognition and image matching. Background technique [0002] Business cards are a medium that people often use to record and transmit contact information. In daily business activities, a large number of business card exchanges are often involved. People pass contact information and establish contact with other people through the exchange of business cards. [0003] The business card recognition technology that has been widely used in mobile phones and microcomputers is to use the principles of image processing and optical character recognition (OCR, Optical Character Recognition) to automatically process and recognize business card images captured by digital cameras or scanned by scanners. Then automatically extract the useful information such as name, phone number, E-mail and so on to add to...

AI Technical Summary

Problems solved by technology

[0004] The business card recognition technology based on the principle of optical character recognition has at least the following three limitations. First, the recognition result is impossible to achieve 100% accuracy, so it is always necessary for the user to check the recognition result before saving the recognized information. Carry out manual inspection and correction to ensure the correctness of the saved information, otherwise, even if there is only one wrong symbol in the saved phone number, E-mail address, website, etc., it will become invalid information; , it is almost impossible to recognize many personalized business cards with complex backgrounds and business cards containing more artistic fonts, even if they can be recognized, the error rate is particularly high; the third is optical character recognition technology and the basic characters of a specific language Therefore, it is necessary to develop a special optical character recognition technology for each language. For example, an optical character recognition technology that only supports Chinese character sets cannot be used to recognize business cards in other languages ​​such as Korean and Japanese.

[0005] In order to solve the above problems, the existing practice in the industry is to transmit the text information that may contain errors obtained through optical character recognition to the server of a specific website, and the server of the website uses these text information that may contain errors to display the text information on the server of the website. Search in the text information database of the pre-stored business card, find out the most similar text information, and replace the text information that may contain errors. Although this is a method to reduce errors, the problem with the above method is that in many cases The server of the website cannot uniquely determine which of the business cards saved by the server of the website, or whether it belongs to the business card saved by the server of the website, based on the text obtained by optical character recognition.

However, under the premise of ensuring the recognition accuracy, the rejection rate should be as low as possible, otherwise the business card recognition will lose its practical value

The above-mentioned rejection rate refers to the number of 100 business card recognition requests submitted to the server of the website, how many times the server of the website tells that it cannot be recognized, and the reason for the inability to recognize may be that the submitted image is not saved in the server of the website The image of the business card taken and the corresponding text information may also be that the server of the website cannot determine which business card is captured by the submitted image, so it refuses to return the corresponding text information

The current image matching algorithm must ensure 100% recognition accuracy, but its rejection rate is usually relatively high

Images