A finger venogram quality evaluation method based on cascade optimization CNN and a device thereof

Abstract

The invention discloses a finger venogram quality evaluation method based on cascade optimization CNN and a device thereof. The gray scale image of finger vein image is sent to the pre-trained cascadeoptimized CNN for feature extraction and classification according to the extracted features and preset quality candidates. The quality of finger vein image is evaluated by using the classification results as the evaluation results. Simplify the final model to achieve efficient and cost-effective quality assessment.

Description

technical field [0001] The invention relates to the field of biometrics, in particular to a method for evaluating the quality of finger vein images based on cascade optimized CNN and a device thereof. Background technique [0002] At present, finger vein recognition, as an emerging biometric identification technology, has been more and more popularized in the application process. Due to different users and devices, the quality of finger vein images collected each time is also high or low. Low-quality finger vein images are usually blurred. If low-quality finger vein images are registered in the database, it will seriously affect later applications. The feature extraction and matching at the time lead to poor recognition performance of the system. Therefore, quality assessment is required after each acquisition to ensure that high-quality finger vein images are registered in the database. In the prior art, CNN (convolutional neural network algorithm) is mostly used to conver...

AI Technical Summary

Problems solved by technology

Due to different users and devices, the quality of finger vein images collected each time is also high or low. Low-quality finger vein images are usually blurred. If low-quality finger vein images are registered in the database, it will seriously affect later applications. The feature extraction and matching at the time lead to poor recognition performance of the system, so quality assessment is required after each acquisition to ensure that high-quality finger vein images are registered in the database

In the prior art, CNN (convolutional neural network algorithm) is mostly used to convert the collected finger vein images into grayscale images and binary images for feature extraction, and to evaluate the extracted features by overlaying, although this method can complete Quality assessment, but setting two separate CNN models for feature extraction not only requires a lot of training overhead, but also takes up a lot of space in the final model, and the existing methods need to preprocess the grayscale image to obtain Binary map, the steps are more complicated

Images