How does the platform verify the authenticity of the photo holding the ID card?

There are two ways to review the handheld ID card photos provided by users, one is manual review, and the other is automatic review by the document review system. Manual review is inefficient, but it is highly reliable. Automatic review This method is highly efficient and can reduce labor costs, but it has poor reliability and cannot accurately identify invalid photos.

Technical implementation elements:

3. The purpose of the present invention is to provide a method and system for identifying handheld ID card photos to improve the current problem that automatic review cannot identify invalid photos.

4. In order to achieve the above-mentioned object of the invention, embodiments of the present invention provide the following technical solution: a method for identifying handheld ID card photos, including the following steps: s10, receiving the handheld ID card photo input by the user; s20, determine whether the resolution of the received handheld ID card photo reaches the set threshold; if yes, proceed to step s30; if not, process the received handheld ID card photo to improve the resolution of the photo, and then proceed to step s30 s30; s30, identify the faces in the handheld ID card photo, obtain the number of recognized faces, and determine whether the handheld ID card photo is qualified based on the number of recognized faces.

5. In a more optimized solution, step s40 is also included. From the recognition result of the face in the hand-held ID card photo in step s30, the face area image is intercepted, and the intercepted person is Perform color recognition on the face area image to determine whether it is the color of a human face. If not, the handheld ID card photo is determined to be unqualified, otherwise it is determined to be qualified.

6. In the optimized solution, in the step s20, the step of processing the received handheld ID card photo to improve the resolution of the photo includes: s201, collecting several real-life scenes. Image; s202, perform downsampling processing on each image to reduce the image resolution. The image before downsampling is used as a high-resolution image h, and the image after downsampling is used as a low-resolution image l. l and h form a set of effective image pair; s203, use the image pair obtained in step s202 to train the neural network model. During training, the low-resolution image l is enlarged and restored to the high-resolution image sr, and then compared with the original high-resolution image h Compare, and the difference is used to adjust the parameters of the neural network model; through iterative training, the final applicable neural network model is obtained; s204, input the handheld ID card photo whose resolution does not reach the set threshold into the final result obtained in step s203. /p>

The neural network model is reconstructed to obtain the handheld ID card photo with improved resolution.

7. In the optimized solution, in step s30, the process of identifying faces in photos of hand-held ID cards includes: s301, collecting several images containing faces, and Extract hog features from each image, and the obtained hog feature descriptors are used as positive samples; s302, collect a number of images that do not contain faces, and extract hog features from each image, and the obtained hog feature descriptors are used as negative samples; s303, Use the support vector machine algorithm to train positive samples and negative samples to obtain the trained two-classification model; s304, use the two-classification model to detect difficult examples on images that do not contain faces, and obtain difficult example samples; s305, detect difficult example samples Extract the hog feature, obtain the hog feature descriptor, and then retrain the two-classification model obtained in step s303, and iterate repeatedly until the end of the training to obtain the final classification model; s306, perform a sliding scan on the different sizes of the hand-held ID card photo, and extract the hog feature , and use the classification model finally trained in step s305 for classification. If the detection is determined to be a human face, then calibrate it.

8. In a more optimized solution, when the amount of collected negative sample data is not enough, use the collected pictures without faces to randomly crop them, and then extract hog features to supplement them.

9. An identification system for handheld ID card photos, including: a data collection module for receiving handheld ID card photos input by the user; a resolution verification module for judging the received handheld ID card photos Whether the resolution reaches the set threshold; if so, the handheld ID card photo is output to the primary recognition module; if not, the received handheld ID card photo is processed to improve the resolution of the photo, and then output to the primary recognition module Module; a primary recognition module, used to recognize the faces in the handheld ID card photo, obtain the number of recognized faces, and determine whether the handheld ID card photo is qualified based on the number of recognized faces.

10. Compared with the existing technology, the method or system of the present invention can effectively solve the problem of unclear interception of ID cards, lack of ID card face in the ID card, national emblem of the ID card in the ID card, etc. There is no holder in the photo, the facial features of the holder are not in the photo, the ID card of the holder is blocked in the photo (the face in the ID card is blocked), the holder is holding a copy of the ID card (black and white), etc. For questions about qualified photos, automatic and quick verification is implemented.

11. This method is applied to the order review system to automatically identify and intercept unqualified photos, and timely feedback to the business front-end that the unqualified pictures avoid users from repeating the submission process after the order review is rejected, improve the user experience, and calibrate user errors in a timely manner. Act with license. At the same time, the return rate is reduced, the same unqualified photo is returned repeatedly, and the cost of communication with the business end is saved.