The intrusion of foreign objects into railways poses a great threat to the reliability and safety of railway systems. In order to effectively avoid such phenomena, this paper studies a method for detecting foreign objects in railway tracks based on deep learning principles. This method mainly includes two parts: rail detection and foreign object recognition. The rail detection part adopts a deep learning method based on UNet network semantic segmentation, which optimizes the convolutional structure of the UNet network into depthwise separable convolutions and inserts a spatial pyramid structure to improve the detection speed The foreign object recognition part adopts YOLOv5's deep learning method, which improves its detection accuracy by changing the loss function to SIoU, and adds a CA attention mechanism to determine the coordinates of the target and recognize and classify foreign objects. The experimental results show that the average recognition accuracy of the algorithm proposed in this paper on the dataset of railway foreign objects reaches 93.7%, which is 5.1% higher than the traditional YOLOv5 algorithm, indicating that the algorithm effectively improves the accuracy of railway foreign object intrusion recognition.
KEYWORDS: Video, Education and training, Feature extraction, Video compression, Deep learning, Video processing, Video acceleration, Data modeling, Databases, Neural networks
This article proposes a no reference video quality assessment method based on deep learning, aiming to simulate human perception of video quality and evaluate videos. This method evaluates the quality of videos by learning effective feature representations in the spatiotemporal domain. First, in the spatial domain, 2D-CNN is used to extract the spatial quality of video frames. Then, in the temporal domain, Recurrent neural network (RNN) and pyramid feature aggregation (PFA) module are used to model the temporal domain and aggregate the frame level feature quality. The experiment shows that the method proposed in this paper has good performance on the KoNViD-1k and CVD2014 datasets, and also indicates that the method has high generalization ability.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.