In today’s era of rapidly advancing biometric technology, Homsh Technology has reached a significant milestone — we have successfully submitted the company’s first patent application for an AI-based iris algorithm. Titled "An Iris Segmentation Method Based on an Improved UNet Model", this patent marks a solid step forward for Hongshi Technology in the in-depth integration of artificial intelligence (AI) and biometrics.
Iris recognition is hailed as the crown jewel of biometric technology. Its uniqueness and stability make it highly favored in high-security scenarios. However, achieving accurate iris region segmentation in complex environments has long been a technical challenge for the industry. Practical challenges such as varying lighting conditions, eyelid occlusion, and motion blur have all tested the robustness of algorithms. After in-depth research, Homsh Technology’s R&D team innovatively combined deep learning technology with iris recognition to create this groundbreaking solution.
The core innovation of this patent lies in the in-depth transformation of the classic UNet network architecture. The R&D team ingeniously adopted MobileNetV3 as the backbone network for feature extraction and integrated the CBAM (Convolutional Block Attention Module) attention mechanism, enabling the algorithm to automatically focus on the key features of the iris. Meanwhile, by introducing the ASPP (Atrous Spatial Pyramid Pooling) multi-scale dilated convolution module, the algorithm can better capture the global contextual information of images, significantly improving the segmentation accuracy of iris boundaries.
More notably, this patent also innovatively integrates dual-band liveness detection technology. Through the coordinated collection of two near-infrared bands (810nm and 940nm), the system can leverage the difference in hemoglobin absorption across different bands to effectively distinguish between real human eyes and spoofing attacks such as photos and videos. This design not only enhances the accuracy of iris recognition but, more importantly, improves its security. Test data shows that the system’s false acceptance rate (FAR) and false rejection rate (FRR) have been significantly optimized, fully meeting the strict requirements of the ISO 19794-6 international standard.
In terms of technical implementation, the R&D team also fully considered the needs of practical applications. By adopting advanced optimization strategies such as mixed-precision training and cosine annealing learning rate scheduling, they not only accelerated the model training process but also significantly improved the algorithm’s generalization ability. The system supports ONNX format conversion and can be flexibly deployed on various hardware platforms, truly achieving a seamless transition from laboratory research to industrial application.
The submission of this patent not only demonstrates Homsh Technology’s technical strength in AI algorithm R&D but also highlights the company’s determination to pursue continuous innovation in the field of iris recognition. From traditional image processing methods to the full application of deep learning technology, and from a single recognition function to comprehensive security protection integrated with liveness detection, Homsh Technology is driving technological progress across the entire industry through concrete actions.
Looking ahead, Homsh Technology will continue to deepen its research in iris recognition technology. Taking this patent as a starting point, we will constantly explore new frontiers in the integration of AI and biometric technology. We believe that through sustained technological innovation and the accumulation of intellectual property rights, Homsh Technology will surely deliver safer and more convenient identity authentication experiences to users in broader application scenarios such as intelligent security, financial payment, and smart cities.
The submission of this patent is not only an important milestone in Homsh Technology’s development journey but also a new starting point for us to move toward our vision of "becoming a world-leading provider of iris recognition technology". We would like to thank every partner for their trust and support. Homsh Technology will continue to be driven by innovation, safeguard security with technology, and make identification smarter.
