Dim Target Detection Method Based on Deep Learning in Complex Traffic Environment

Although the current vehicle detection and recognition framework based on deep learning has its own characteristics advantages, it is difficult to effectively combine multi-scale multi category features, there still room for improvement in performance. Based this, an improved fast R-CNN convolutional neural network proposed detect dim targets complex traffic environment. The model of introduced into image environment, a structure optimization method proposed, which replaces VGG16 RCNN with RESNET make suitable small target background. Max pooling down sampling method, then feature pyramid RPN generate candidate box optimize network. After training 1497 images, environment images are identified tested. results show that accuracy better than other comparison methods, highest 94.7%.