基于多模型协作和反向校验的快速指纹图像检索研究

Research on Fast Fingerprint Image Retrieval Based on Multi Model Collaboration and Reverse Validation

在生物特征识别领域,指纹识别因其独特性被广泛应用。随着指纹识别场景的普及,逐一匹配指纹不具有实际意义。因此,研究一种快速高效的指纹图像检索识别技术具有重要意义。本文旨在研究指纹细节点的快速检索和精细匹配问题,提出了一种快速高效的指纹检索模型。该模型结合基于区域的细节点特征匹配算法减少噪声干扰,设计粗粒度残差网络提取数据点特征,此外,利用基于卡尔曼滤波的改进遗传算法二次筛选数据样本,最后,利用BEBLID描述符对样本数据进行精准匹配。经过综合分析,本文所提出的模型运行时间较快,内存消耗量较小。

In the field of biometric recognition,fingerprint identification are widely used due to their uniqueness.With the popularization of fingerprint recognition scenarios,matching fingerprints one by one is not of practical significance.Therefore,studying a fast and efficient fingerprint image retrieval and recognition technology is of great significance.This article aims to study the problem of fast retrieval and fine matching of fingerprint details,and proposes a fast and efficient fingerprint retrieval model.This model combines a region based detail point feature matching algorithm to reduce noise interference,designs a coarse-grained residual network to extract data point features,and uses an improved genetic algorithm based on Kalman filtering to screen data samples twice.Finally,the BEBLID descriptor is used to accurately match the sample data.After comprehensive analysis,the model proposed in this article has a faster running time and lower memory consumption.