Urban subsurface infrastructures, e.g., pipelines and roads, are aging with the expansion of modern cities. Benefiting from the capability of nondestructive detection, ground penetrating radar (GPR) has been widely applied to underground objects or disasters detection, and GPR B-scan images are employed by manual interpretation. While, this way of high subjectivity and uncertainty inevitably results in failure of detection. Meanwhile, the shortage of labelled images greatly impedes the automatization and intelligentization of underground disaster detection based on GPR. Many data simulation techniques, e.g., forward modelling, were used to augment images for training; however, the generated forward images were not similar enough to the real B-scan data, which makes recognition a challenging task. To address this problem, we proposed a novel B-scan image simulation method based on generative adversarial network to generate synthetic images for training detection networks. Our network utilizes DenseNet as the backbone network of generator to extract image features, and a weighted total variation regularization term to regularize the loss function of the network. The comparison and ablation experiments verified that our network could generate simulation images with high similarity to real GPR B-scan images. We believe that this work contributes to the intelligent processing and analysis of GPR data, and improves the efficiency of underground disaster detection.
Semantic segmentation of large-scale point clouds is a research hotspot in aerospace remote sensing technology. Large-scale point clouds face the curse of dimensionality, and there are two problems to resolve, i.e., severe feature loss caused by down-sampling and tricky balance between segmentation performance and computational cost. In this paper, we propose the FAR-Net, a novel point cloud segmentation method based on feature aggregation-recoding, to infer point-level semantics of large-scale point clouds at reasonable cost. This network has three major features. First, local difference attention coding preserves feature details. Second, a hierarchical feature aggregation module based on keypoints and voxels is designed with two branches. The keypoint branch uses k-means clustering to generate keypoints that are globally associated by Point Transformer. The voxel branch adopts a multi-scale structure with three cascaded 3D CNN modules to extract local spatial features. Third, a binary feature recoding module divides intra-cluster point clouds into two categories by a binary classifier, and assigns keypoint features and difference features according to the binary label to increase feature differences. Experiments show that our network achieves competitive segmentation performance on both indoor and outdoor large-scale 3D point cloud datasets.
This paper deals with modulation classification under the alpha-stable noise condition. Our goal is to discriminate orthogonal frequency division multiplexing (OFDM) modulation type from single carrier linear digital (SCLD) modulations in this scenario. Based on the new results concerning the generalized cyclostationarity of these signals in alpha-stable noise which are presented in this paper, we construct new modulation classification features without any priori information of carrier frequency and timing offset of the received signals, and use support vector machine (SVM) as classifier to discriminate OFDM from SCLD. Simulation results show that the recognition accuracy of the proposed algorithm can be up to 95% when the mix signal to noise ratio (MSNR) is up to ?1 dB.
In the indoor environment, the Global Positioning System (GPS) cannot provide accurate location information due to the complexity of the internal structure of buildings and the interference of indoor equipments. Pedestrian Dead Reckoning (PDR) is an indoor positioning method that relies on smartphone carrying sensors (such as accelerometer, gyroscope, and magnetometer) without requiring additional equipments. The intense electromagnetic interferences caused by indoor electronic equipments have serious influences on the accuracies of indoor location. This paper presents an improved step detection and counting method based on PDR. There is continuity in the walking process. In order to make full use of this feature, this paper firstly filters and differential processes the acceleration data obtained from the smartphone, then uses the improved peak detection algorithm to detect and count the steps. This paper compares two algorithms of step length estimation, and presents an improved step length estimation algorithm based on differential acceleration data. As a result, this paper improves the accuracies of both step number detection and step length estimation.
Abstract The explosion of wireless technology has made it a hot topic in undergraduate education. Many undergraduate students are intrigued by the secrets behind wireless communication and networking, but few institutions can provide hands-on laboratories in their networking courses due to expensive hardware equipment. Funded by a collaborative NSF TUES type II project, a series of affordable and evolvable software defined radio (SDR) based laboratories was implemented and institutionalized at three institutions to demonstrate its capability and adaptability. As a participating institution, Central State University worked closely with Wright State University and Miami University and successfully adapted the novel SDR based laboratories. We further initialized our own laboratory modules to improve undergraduate students' understanding and learning. The laboratory modules were integrated into two undergraduate level networking-related courses, and the course assessment showed positive learning outcomes. The exploratory project is a work in progress and we will continue the development in order to lead a national model of SDR laboratory based courses.
In this paper, we focus on the age of information (AoI) aware intelligent architecture of emergency service by caching placement optimization on satellite. Such an intelligent architecture can effectively improve the AoI of users when the cellular network are broken in a disaster area. Unlike the traditional method, our architecture can learn through interaction with the environment, continuously optimizing the cache policy to improve cache hit rate and system performance. Specifically, we first model the emergency service process of the satellite as a Markov process. Then, we apply the advantage actor-critic (A2C) algorithm to obtain the optimal satellite cache strategy. By extensive simulation results, we can verify that the proposed intelligent architecture outperforms the existing traditional methods.
Deep learning (DL)-based modulation recognition methods are challenging in the case of few labeled samples and underwater impulsive noise. In this letter, we propose a novel network structure named IAFNet to achieve higher recognition accuracy of modulation signals with fewer samples in underwater impulsive noise environment. The IAFNet integrates impulsive noise preprocessing (INP), attention network (AN) and few-shot learning (FSL) to extract features more effectively through denoising and task-driven. Experimental results on simulation and practical data show that the IAFNet attains stronger anti-noise performance and better recognition performance on fewer labeled samples. Compared with other methods, the classification accuracy is improved by about 7%.
This paper presents a new region-based unified tensor level set model for image segmentation. This model introduces a three-order tensor to comprehensively depict features of pixels, e.g., gray value and the local geometrical features, such as orientation and gradient, and then, by defining a weighted distance, we generalized the representative region-based level set method from scalar to tensor. The proposed model has four main advantages compared with the traditional representative method as follows. First, involving the Gaussian filter bank, the model is robust against noise, particularly the salt- and pepper-type noise. Second, considering the local geometrical features, e.g., orientation and gradient, the model pays more attention to boundaries and makes the evolving curve stop more easily at the boundary location. Third, due to the unified tensor pixel representation representing the pixels, the model segments images more accurately and naturally. Fourth, based on a weighted distance definition, the model possesses the capacity to cope with data varying from scalar to vector, then to high-order tensor. We apply the proposed method to synthetic, medical, and natural images, and the result suggests that the proposed method is superior to the available representative region-based level set method.
Target detection and tracking algorithms are one of the key technologies in the field of autonomous driving in intelligent transportation, providing important sensing capabilities for vehicle localization and path planning. Siamese network-based trackers formulate the visual tracking mission as an image-matching process by regression and classification branches, which simplifies the network structure and improves the tracking accuracy. However, there remain many problems, as described below. (1) The lightweight neural networks decrease the feature representation ability. It is easy for the tracker to fail under the disturbing distractors (e.g., deformation and similar objects) or large changes in the viewing angle. (2) The tracker cannot adapt to variations of the object. (3) The tracker cannot reposition the object that has failed to track. To address these issues, we first propose a novel match filter arbiter based on the Euclidean distance histogram between the centers of multiple candidate objects to automatically determine whether the tracker fails. Secondly, the Hopcroft–Karp algorithm is introduced to select the winners from the dynamic template set through the backtracking process, and object relocation is achieved by comparing the Gradient Magnitude Similarity Deviation between the template and the winners. The experiments show that our method obtains better performance on several tracking benchmarks, i.e., OTB100, VOT2018, GOT-10k, and LaSOT, compared with state-of-the-art methods.
The mainstream methods for change detection in synthetic-aperture radar (SAR) images use difference images to define the initial change regions. However, methods can suffer from semantic collapse, which makes it difficult to determine semantic information about the changes. In this paper, we proposed a hierarchical fusion SAR image change-detection model based on hierarchical fusion conditional random field (HF-CRF). This model introduces multimodal difference images and constructs the fusion energy potential function using dynamic convolutional neural networks and sliding window entropy information. By using an iterative convergence process, the proposed method was able to accurately detect the change-detection regions. We designed a dynamic region convolutional semantic segmentation network with a two-branch structure (D-DRUNet) to accomplish feature fusion and the segmentation of multimodal difference images. The proposed network adopts a dual encoder–single decoder structure where the baseline is the UNet network that utilizes dynamic convolution kernels. D-DRUNet extracts multimodal difference features and completes semantic-level fusion. The Sobel operator is introduced to strengthen the multimodal difference-image boundary information and construct the dynamic fusion pairwise potential function, based on local boundary entropy. Finally, the final change result is stabilized by iterative convergence of the CRF energy potential function. Experimental results demonstrate that the proposed method outperforms existing methods in terms of the overall number of detection errors, and reduces the occurrence of false positives.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)