Research on low-light face detection method based on Faster R-CNN
With the development of deep learning,face detection algorithms have reached an excellent level of accuracy and speed in ideal environments.However,in real-world scenarios,it is difficult to achieve the ideal detection environment,and factors such as low lighting and blurred images can cause interference,leading to a decrease in the accuracy of face detection.To address this issue,a face detection method based on an improved Faster-RCNN combined with the MSRCR algorithm is proposed.Firstly,the MSRCR algorithm is used to perform automatic white balance(AWB)on the images to address the issues of low lighting and blurring.Finally,Soft NMS is employed to optimize the candidate box selection in Faster-RCNN,solving the problem of missing overlapping detection targets.Experimental results show that the improved Faster R-CNN combined with the MSRCR algorithm has higher detection accuracy compared to traditional face detection methods in low-light or blurred conditions.
face detectiondeep learningimage enhancementlow illumination
万方数据
