基于特征分块的赤足足迹人身识别算法

为了提高赤足足迹人身识别算法的准确率，本文提出了一种基于深度学习的足迹识别算法。由于足底各区域压力的不同导致了赤足足迹各部分包含的信息量存在一定的差异性，为了获取更稳定、区分度更高的特征，采用ResNet50作为基础网络，在特征层进行分块处理。本文基于2000人的赤足足迹库进行训练，利用500人1000幅测试图在3000人的赤足测试库上进行测试。所提出算法的首位识别准确率达到了98.50%，优于常规的ResNet50网络。实验证明，本文提出的基于特征分块的足迹识别算法在赤足足迹识别中获得了很好的识别效果。

Research on Barefoot Footprint Person Identification Algorithm Based on Feature Partitioning

To improve the accuracy of footprint personal identification, a footprint recognition algorithm based on deep learning was proposed. Due to the different pressure in each area of the plantar, the information contained in each part of the footprint is different to some extent. In order to obtain more stable and distinguishable features, ResNet50 was used as the basic configuration and block processing was carried out in the feature layer. This paper conducted training on the barefoot footprint database based on 2000 people. The test was carried out on a barefoot test set of 3000 people using 1000 test images from 500 people. The first recognition accuracy of the proposed algorithm was 98.50%, which was better than the conventional ResNet50 network. Experimental results show that the proposed algorithm based on feature partitioning achieved a good recognition effect.