Research On Visual Comfort Lighting Control System Based On Iris Segmentatio

The iris is the circular membrane located between the black pupil and the white sclera in the human eye.With its safe,unique,and stable characteristics,it plays a key role in the fields of iris recognition,gaze estimate,and visual health.The performance of the current iris segmentation algorithm will be affected under non-limiting conditions,especially in low-quality iris images such as low pixels,blurred imaging,uneven lighting,and eyelid occlusion.This dissertation proposed the segmentation algorithm SERes Next-UNet(SRN-UNet)suitable for low-quality iris images,combined proposed algorithm with the evaluation algorithm of human eye visual comfort to build an adaptive visual comfort lighting control system,conducted the deployment and functional test of the system based on the Jetson AGX Xavier embedded platform.The main work includes:Firstly,the research on iris segmentation algorithm was summarized,and the segmentation algorithm SRN-UNet suitable for low-quality iris images was proposed based on the U-Net.In the encoding stage,the SE-Res Next module was added,which cascaded the SENet(Squeeze-and-Excitation Network)module after the Rex Next module.The SE-Res Next module can focus on key features and improve the segmentation accuracy and network performance without increasing network parameters.In the decoding stage,part of the convolutional layer was removed to improve training speed.The test results showed that proposed algorithm outperformed all tested algorithm based on the self-built dataset,the Miou,F1 Score,and Precision of the proposed algorithm reached 95.19%,97.48%,and 97.82% respectively.The running speed was faster than the U-Net.Secondly,in order to alleviate the visual burden of VDT workers brought by the long-term use of computer monitors,based on proposed iris segmentation algorithm,this dissertation combined with fuzzy relationship between pupil diameter variation and LED light brightness control signal under different visual comfort levels to build a visual comfort lighting control system that can adaptively adjust the ambient illumination.The system calculated relative variation change of pupil diameter through the human eye detection algorithm of YOLOv4,SRN-UNet iris segmentation algorithm and related equations.Then,the relative change of pupil diameter was fed to fuzzy controller to obtain the LED light brightness control signal to realize the adaptive adjustment of the ambient illumination.Finally,the visual comfort lighting control system was deployed on the Jetson AGX Xavier embedded platform,and the software functions and the overall system functions were tested and analyzed.The results showed that the entire visual comfort lighting control system can adaptively adjust the environmental illuminance according to the visual comfort state of VDT workers,which can effectively reduce the visual burden of VDT workers and create a good lighting environment.