Research on the Face Detection of Vehicle Recorder in Low-lighting EnvironmentResearch on the Face Detection of Vehicle Recorder in Low-lighting Environment
Qian-yi Li 1, Yuan FANG 2,*, Xue-Ping Wang 1 , Wei-zhen Wang 3 , Nian-yu Zou 1 (1:School of Information Science and Engineering, Dalian Polytechnic University, China 2:School of Engineering Practice and Innovation-Entrepreneurship Education, Dalian Polytechnic University, China 3: Human Factors and Intelligent Design Research Center, Dalian Polytechnic University, China)
Abstract—Face detection has been a research direction of great interest in the field of deep learning. In vehicle recorders, face detection in low-lighting environments has been a challenging problem which leads to varying face detection accuracy. Towards the improving face detection accuracy in low-lighting environments, a face fusion-based image enhancement method is proposed, which employs a combination of image enhancement and deep learning methods. The method includes a low-lighting enhancement module and a face detection module. The low-lighting enhancement module, which performs exposure enhancement of the captured image by improving Retinex network to improve the image quality and color. The face detection module performs the detection by target detection method, which utilizes a lightweight approach to alleviate the model training process. The model in this paper has a detection accuracy of more than 95% in low-lighting environments, and the quality of the exposed images is evaluated by objective assessment methods, and the quality of their enhanced images are all improved. Keywords- face detection; low-lighting environment; image enhancement; deep learning;YOLO |