Facial Recognition Technology for Scenic Spot Monitoring Based on U2Net and FFC

Abstract

To ensure the safety of scenic spots and achieve intelligent management of scenic spots, a face recognition method based on U2Net and FFC is proposed to achieve monitoring face recognition under different occlusion conditions. It consists of a small area regular occlusion face recognition model and a large area irregular occlusion face recognition model. Firstly, a face recognition model grounded on an improved residual network-U2Net is raised to address the problem of small area rule occlusion. This model combines a global convolution module, a feature pyramid network, and a mask learning unit. When evaluating facial recognition methods, multiple evaluation metrics were used, including recognition accuracy, F1-score, recognition rate, structural similarity index, peak signal-to-noise ratio, learning perceptual image block similarity, and Frecht approximation distance. These indicators evaluate the performance of the model under small and large area irregular occlusion conditions from different perspectives, ensuring the comprehensiveness and reliability of the evaluation. The findings denote that the average recognition accuracy of the enhanced residual network-U2Net is as high as 98.7%, the average F1-score is 0.983, and the average recognition rate is 99.5%. Secondly, in response to the problem of large-scale irregular occlusion in facial recognition, a fast Fourier convolution generative adversarial network is proposed, which combines generative adversarial network and Fourier feature convolution to repair and recognize facial images. The outcomes denote that the average structural similarity index and peak signal-to-noise ratio of the model are 0.878 and 34.7dB, respectively, and the average accuracy and recognition rate are 91.0% and 92.6%, respectively. The above results denote that the proposed facial recognition method exhibits superior performance under different occlusion conditions and can effectively promote the intelligent development of scenic area management.