Road recognition from optical remote sensing images is important for many applications like intelligent transportation system. Currently, convolutional neural networks (CNNs) methods are widely utilized in road recognition. However, many CNN methods hardly get good recognition performance when they process high-resolution images with large road width variance and complex background. For this problem, we develop a road recognition method based on a multi-scale convolutional network (MCN) with multi-level feature fusion. MCN comprises several CNNs with different scales of inputs and thus can extract multi-scale features. Each CNN fuses low-level geometrical features, middle-level features and high-level semantic features respectively from shallow, middle and deep layers. The multi-scale scheme and multi-level feature fusion make the MCN capable to handle large road width variance and complex background. Our method is validated on a manually labeled visible remote sensing image dataset. Moreover, our method is compared with CNNs without multi-scale or multi-level feature fusion and a fully convolutional network (FCN). The experimental results show that our method can well deal with complex visible remote sensing images with large road width variance.
The study of plants in space plays an important role in serving astronauts. Automatic segmentation of space plant image (SPI) provides an effective method for studying plants, and many plant segmentation methods have been proposed. However, segmentation of SPI is still challenging. Because the number of SPI is small, which greatly increases the difficulty in model training (especially deep-network-based model). For dealing with this problem, we propose a plant segmentation method based on a generative adversarial network. Our method consists of a generative network (GN) and a discriminant network (DN). The GN firstly extracts features from an input image, and then generates a feature map by developing multiple convolution and deconvolution layers. The DN merges the feature map with an actual plant image, and then computes a segmentation result by a deep convolutional network. In DN, the addition of the feature map improves segmentation accuracy of DN, and reduces the requirements of training images during training DN. Several experiments are made, and the experimental results show that our method performs well when a small number of training images is provided for model training.
System logs record the daily status of operating systems, application software, firewalls, etc. Analyzing system logs can help to prevent and eliminate information security events in real time. In this paper, we propose to analyze the system logs for anomalous event detection based on natural language processing. First, we use doc2vec of natural language processing algorithm to construct sentence vectors, then apply several state-of-the-art classification algorithms on the sentence vectors for anomaly detection. The system logs generated by the Thunderbird supercomputer are adopted here to verify the proposed method. The results show that doc2vec combined with machine learning classification algorithms could not only effectively extract the semantic information of the logs, but also perform excellent anomaly detection.
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.