Face recognition base library optimization method, system, equipment and readable storage medium

Abstract

The invention discloses a face recognition base library optimization method, a face recognition base library optimization system, face recognition base library optimization equipment and a readable storage medium. The face recognition base library optimization method comprises the steps of: classifying base library pictures through using a classifier when constructing a base library, and carryingout credibility grading on the base library pictures; collecting a new picture, performing similarity detection on the new picture and first-level credibility pictures, updating the new picture into the base library when the similarity is greater than a first threshold value, and performing classification; when the similarity is smaller than the first threshold value, conducting similarity detection on the new picture and second-level credibility pictures, and updating the new picture into the base library when a numerical value of the similarity is larger than a second threshold value, and performing classification; and ending the task when the similarity between the new picture and the second-level credibility pictures is smaller than the second threshold value. According to the face recognition base library optimization method, the face recognition base library optimization system and the face recognition base library optimization equipment, the accuracy of the face recognition algorithm in application is improved, and compared with an existing method, the face recognition accuracy can be effectively improved at low cost without retraining the model, using a special camera and performing manual intervention.

Description

technical field [0001] The present invention relates to the technical field of image recognition, in particular to a method, system, device, and readable storage medium for optimizing a face recognition base database. Background technique [0002] Face recognition is a biometric technology for identification based on human facial feature information. Most of the current face recognition algorithms use the method of converting face pictures into feature vectors and then calculating distances for face recognition, that is, face comparison is performed based on the similarity of two pictures. This method has certain limitations in the application of the real environment. Just like a person, the face images collected under different lighting, different angles, and different times will have certain differences. The occlusion and so on will also lead to a decrease in the recognition rate and accuracy of the algorithm. For addressing the above problem, existing solution has follo...

AI Technical Summary

Problems solved by technology

[0003] One solution is to retrain the model for the scene, but this method is limited by the amount of data in a specific scene, and the model itself is also a difficulty in recognizing some poor-quality pictures, that is, it is difficult to meet requirements such as side faces, lighting, etc. Under certain conditions, it is also difficult to update in real time

The second solution is to improve the quality of captured pictures by collecting 3D images, or using structured light and other technologies, but this method has high requirements for hardware, and usually requires a dedicated camera or multiple cameras to achieve

The third solution is to manually update or expand the bottom library pictures. This method requires a certain amount of labor costs. At the same time, it is difficult to achieve comprehensive and timely updates when the amount of data is large.

Images