scholarly journals Direct Aerial Visual Geolocalization Using Deep Neural Networks

2021 ◽  
Vol 13 (19) ◽  
pp. 4017
Author(s):  
Winthrop Harvey ◽  
Chase Rainwater ◽  
Jackson Cothren
Keyword(s):  
Neural Networks ◽  
Deep Neural Networks ◽  
Ground Level ◽  
Satellite System ◽  
Input Image ◽  
Average Error ◽  
Washington County ◽  
Localization Accuracy ◽  
Landscape Information ◽  
Global Navigation Satellite

Unmanned aerial vehicles (UAVs) must keep track of their location in order to maintain flight plans. Currently, this task is almost entirely performed by a combination of Inertial Measurement Units (IMUs) and reference to GNSS (Global Navigation Satellite System). Navigation by GNSS, however, is not always reliable, due to various causes both natural (reflection and blockage from objects, technical fault, inclement weather) and artificial (GPS spoofing and denial). In such GPS-denied situations, it is desirable to have additional methods for aerial geolocalization. One such method is visual geolocalization, where aircraft use their ground facing cameras to localize and navigate. The state of the art in many ground-level image processing tasks involve the use of Convolutional Neural Networks (CNNs). We present here a study of how effectively a modern CNN designed for visual classification can be applied to the problem of Absolute Visual Geolocalization (AVL, localization without a prior location estimate). An Xception based architecture is trained from scratch over a >1000 km2 section of Washington County, Arkansas to directly regress latitude and longitude from images from different orthorectified high-altitude survey flights. It achieves average localization accuracy on unseen image sets over the same region from different years and seasons with as low as 115 meters average error, which localizes to 0.004% of the training area, or about 8% of the width of the 1.5 × 1.5km input image. This demonstrates that CNNs are expressive enough to encode robust landscape information for geolocalization over large geographic areas. Furthermore, discussed are methods of providing uncertainty for CNN regression outputs, and future areas of potential improvement for use of deep neural networks in visual geolocalization.

scholarly journals Latent Multi-Task Architecture Learning

2019 ◽  
Vol 33 ◽  
pp. 4822-4829 ◽  
Author(s):  
Sebastian Ruder ◽  
Joachim Bingel ◽  
Isabelle Augenstein ◽  
Anders Søgaard
Keyword(s):  
Neural Networks ◽  
Recent Work ◽  
Deep Neural Networks ◽  
Synthetic Data ◽  
Approaches To Learning ◽  
Average Error ◽  
Task Learning ◽  
Task Architecture ◽  
Parameter Sharing

Multi-task learning (MTL) allows deep neural networks to learn from related tasks by sharing parameters with other networks. In practice, however, MTL involves searching an enormous space of possible parameter sharing architectures to find (a) the layers or subspaces that benefit from sharing, (b) the appropriate amount of sharing, and (c) the appropriate relative weights of the different task losses. Recent work has addressed each of the above problems in isolation. In this work we present an approach that learns a latent multi-task architecture that jointly addresses (a)–(c). We present experiments on synthetic data and data from OntoNotes 5.0, including four different tasks and seven different domains. Our extension consistently outperforms previous approaches to learning latent architectures for multi-task problems and achieves up to 15% average error reductions over common approaches to MTL.

scholarly journals Performance Evaluation of Pedestrian Locations Based on Contemporary Smartphones

2017 ◽  
Vol 2017 ◽  
pp. 1-10 ◽  
Author(s):  
Jalal Ibrahim Al-Azizi ◽  
Helmi Zulhaidi Mohd Shafri
Keyword(s):  
Performance Evaluation ◽  
Environmental Factors ◽  
Global Navigation Satellite System ◽  
Low Cost ◽  
Satellite System ◽  
Navigation Satellite ◽  
Localization Accuracy ◽  
Environment Type ◽  
Global Navigation ◽  
Global Navigation Satellite

Nowadays, a Global Navigation Satellite System (GNSS) unit is embedded in nearly every smartphone. This unit allows a smartphone to detect the user’s location and motion, and it makes functions, such as navigation, tracking, and compass applications, available to the user. Therefore, the GNSS unit has become one of the most important features in modern smartphones. However, because most smartphones incorporate relatively low-cost GNSS chips, their localization accuracy varies depending on the number of accessible GNSS satellites, and it is highly dependent on environmental factors that cause interference such as forests and buildings. This research evaluated the performance of the GNSS units inside two different models of smartphones in determining pedestrian locations in different environments. The results indicate that the overall performances of the two devices were related directly to the environment, type of smartphone/GNSS chipset, and the application used to collect the information.

scholarly journals How Many Features is an Image Worth? Multi-Channel CNN for Steering Angle Prediction in Autonomous Vehicles

2021 ◽  
Author(s):  
Jason Munger ◽  
Carlos W. Morato
Keyword(s):  
Neural Networks ◽  
Autonomous Vehicles ◽  
Deep Neural Networks ◽  
Spatial Information ◽  
Image Data ◽  
Input Image ◽  
Steering Angle ◽  
Rgb Images ◽  
And Performance ◽  
The Individual

This project explores how raw image data obtained from AV cameras can provide a model with more spatial information than can be learned from simple RGB images alone. This paper leverages the advances of deep neural networks to demonstrate steering angle predictions of autonomous vehicles through an end-to-end multi-channel CNN model using only the image data provided from an onboard camera. Image data is processed through existing neural networks to provide pixel segmentation and depth estimates and input to a new neural network along with the raw input image to provide enhanced feature signals from the environment. Various input combinations of Multi-Channel CNNs are evaluated, and their effectiveness is compared to single CNN networks using the individual data inputs. The model with the most accurate steering predictions is identified and performance compared to previous neural networks.

scholarly journals Signal Extraction from GNSS Position Time Series Using Weighted Wavelet Analysis

2020 ◽  
Vol 12 (6) ◽  
pp. 992 ◽  
Author(s):  
Kunpu Ji ◽  
Yunzhong Shen ◽  
Fengwei Wang
Keyword(s):  
Time Series ◽  
Wavelet Analysis ◽  
Satellite System ◽  
Signal Extraction ◽  
Average Error ◽  
North East ◽  
Observation Network ◽  
Position Time Series ◽  
Global Navigation Satellite ◽  
Gnss Position Time Series

The daily position time series derived by Global Navigation Satellite System (GNSS) contain nonlinear signals which are suitably extracted by using wavelet analysis. Considering formal errors are also provided in daily GNSS solutions, a weighted wavelet analysis is proposed in this contribution where the weight factors are constructed via the formal errors. The proposed approach is applied to process the position time series of 27 permanent stations from the Crustal Movement Observation Network of China (CMONOC), compared to traditional wavelet analysis. The results show that the proposed approach can extract more exact signals than traditional wavelet analysis, with the average error reductions are 13.24%, 13.53% and 9.35% in north, east and up coordinate components, respectively. The results from 500 simulations indicate that the signals extracted by proposed approach are closer to true signals than the traditional wavelet analysis.

scholarly journals Robust Inter-Vehicle Distance Measurement Using Cooperative Vehicle Localization

Sensors ◽  
2021 ◽  
Vol 21 (6) ◽  
pp. 2048
Author(s):  
Faan Wang ◽  
Weichao Zhuang ◽  
Guodong Yin ◽  
Shuaipeng Liu ◽  
Ying Liu ◽  
...  
Keyword(s):  
Satellite System ◽  
Superior Performance ◽  
Estimation Methods ◽  
Cooperative Localization ◽  
Absolute Position ◽  
Localization Accuracy ◽  
Vehicle Localization ◽  
Communication Devices ◽  
Positioning Method ◽  
Global Navigation Satellite

Precise localization is critical to safety for connected and automated vehicles (CAV). The global navigation satellite system is the most common vehicle positioning method and has been widely studied to improve localization accuracy. In addition to single-vehicle localization, some recently developed CAV applications require accurate measurement of the inter-vehicle distance (IVD). Thus, this paper proposes a cooperative localization framework that shares the absolute position or pseudorange by using V2X communication devices to estimate the IVD. Four IVD estimation methods are presented: Absolute Position Differencing (APD), Pseudorange Differencing (PD), Single Differencing (SD) and Double Differencing (DD). Several static and dynamic experiments are conducted to evaluate and compare their measurement accuracy. The results show that the proposed methods may have different performances under different conditions. The DD shows the superior performance among the four methods if the uncorrelated errors are small or negligible (static experiment or dynamic experiment with open-sky conditions). When multi-path errors emerge due to the blocked GPS signal, the PD method using the original pseudorange is more effective because the uncorrelated errors cannot be eliminated by the differential technique.

scholarly journals Comparative Analysis of Deep Neural Networks for the Detection and Decoding of Data Matrix Landmarks in Cluttered Indoor Environments

2021 ◽  
Vol 103 (1) ◽  
Author(s):  
Tiago Almeida ◽  
Vitor Santos ◽  
Oscar Martinez Mozos ◽  
Bernardo Lourenço
Keyword(s):  
Neural Networks ◽  
Deep Learning ◽  
Deep Neural Networks ◽  
Input Image ◽  
Automated Guided Vehicles ◽  
Data Matrix ◽  
Single Shot ◽  
Indoor Environments ◽  
Data Set ◽  
Valid Solution

AbstractData Matrix patterns imprinted as passive visual landmarks have shown to be a valid solution for the self-localization of Automated Guided Vehicles (AGVs) in shop floors. However, existing Data Matrix decoding applications take a long time to detect and segment the markers in the input image. Therefore, this paper proposes a pipeline where the detector is based on a real-time Deep Learning network and the decoder is a conventional method, i.e. the implementation in libdmtx. To do so, several types of Deep Neural Networks (DNNs) for object detection were studied, trained, compared, and assessed. The architectures range from region proposals (Faster R-CNN) to single-shot methods (SSD and YOLO). This study focused on performance and processing time to select the best Deep Learning (DL) model to carry out the detection of the visual markers. Additionally, a specific data set was created to evaluate those networks. This test set includes demanding situations, such as high illumination gradients in the same scene and Data Matrix markers positioned in skewed planes. The proposed approach outperformed the best known and most used Data Matrix decoder available in libraries like libdmtx.

scholarly journals Using Geomatic Techniques to Estimate Volume–Area Relationships of Watering Ponds

2021 ◽  
Vol 10 (8) ◽  
pp. 502
Author(s):  
Ubaldo Marín-Comitre ◽  
Álvaro Gómez-Gutiérrez ◽  
Francisco Lavado-Contador ◽  
Manuel Sánchez-Fernández ◽  
Alberto Alfonso-Torreño
Keyword(s):  
Storage Capacity ◽  
Laser Scanner ◽  
Satellite System ◽  
Volume Estimation ◽  
Quadratic Functions ◽  
Average Error ◽  
Power Relationship ◽  
Hydrological Simulation ◽  
Stereo Photogrammetry ◽  
Global Navigation Satellite

Watering ponds represent an important part of the hydrological resources in some water-limited environments. Knowledge about their storage capacity and geometrical characteristics is crucial for a better understanding and management of water resources in the context of climate change. In this study, the suitability of different geomatic approaches to model watering pond geometry and estimate pond-specific and generalized volume–area–height (V–A–h) relationships was tested. Terrestrial structure-from-motion and multi-view-stereo photogrammetry (SfM-MVS), terrestrial laser scanner (TLS), laser-imaging detection and ranging (LIDAR), and aerial SfM-MVS were tested for the emerged terrain, while the global navigation satellite system (GNSS) was used to survey the submerged terrain and to test the resulting digital elevation models (DEMs). The combined use of terrestrial SfM-MVS and GNSS produced accurate DEMs of the ponds that resulted in an average error of 1.19% in the maximum volume estimation, comparable to that obtained by the TLS+GNSS approach (3.27%). From these DEMs, power and quadratic functions were used to express pond-specific and generalized V–A–h relationships and checked for accuracy. The results revealed that quadratic functions fit the data particularly well (R2 ≥ 0.995 and NRMSE < 2.25%) and can therefore be reliably used as simple geometric models of watering ponds in hydrological simulation studies. Finally, a generalized V–A power relationship was obtained. This relationship may be a valuable tool to estimate the storage capacity of other watering ponds in comparable areas in a context of data scarcity.

scholarly journals Application of ALS data for GNSS terrain obstacles inventory

2018 ◽  
Vol 63 ◽  
pp. 00014
Author(s):  
Renata Pelc-Mieczkowska ◽  
Dariusz Tomaszewski ◽  
Karolina Jurgielewicz
Keyword(s):  
Laser Scanning ◽  
Reference Model ◽  
Data File ◽  
Satellite System ◽  
Tree Canopy ◽  
Average Error ◽  
Test Models ◽  
Differential Positioning ◽  
Current Constellation ◽  
Global Navigation Satellite

Current constellation of global navigation satellite system (GNSS) ensures signal availability even in severe observational conditions like urban canyon or under tree canopy. However, positioning in such environment remains a challenge because obstacles can block, reflect and diffract GNSS signals which significantly affects accuracy. Those errors are strongly sight dependent and cannot be mitigated in differential positioning that is why, knowledge of the shape and spatial distribution of terrain obstacles is essential. In this paper using of airborne laser scanning (ALS) data for terrain obstacles inventory is presented and evaluated. In proposed method terrain obstacle models have been derived from ASCII ALS data file using open source QGIS with LAStools software suite and dedicated ALSObstModel plugin. Test models were developed for three geodetic control points with different environmental characteristics. For each point reference model from direct tachometry measurements have been obtained. An average error in determining the elevation of the terrain obstacles from ALS based models was 0.6° to 1.7°. This distance corresponds to 3 to 6 minutes of satellite in orbit.

scholarly journals Distributed Localization with Complemented RSS and AOA Measurements: Theory and Methods

2019 ◽  
Vol 10 (1) ◽  
pp. 272 ◽  
Author(s):  
Slavisa Tomic ◽  
Marko Beko ◽  
Luís M. Camarinha-Matos ◽  
Luís Bica Oliveira
Keyword(s):  
Integrated Circuit ◽  
Urban Areas ◽  
Rapid Development ◽  
Satellite System ◽  
Alternative Methods ◽  
Future Research ◽  
Angle Of Arrival ◽  
Localization Accuracy ◽  
Distributed Solutions ◽  
Global Navigation Satellite

Remarkable progress in radio frequency and micro-electro-mechanical systems integrated circuit design over the last two decades has enabled the use of wireless sensor networks with thousands of nodes. It is foreseen that the fifth generation of networks will provide significantly higher bandwidth and faster data rates with potential for interconnecting myriads of heterogeneous devices (sensors, agents, users, machines, and vehicles) into a single network (of nodes), under the notion of Internet of Things. The ability to accurately determine the physical location of each node (stationary or moving) will permit rapid development of new services and enhancement of the entire system. In outdoor environments, this could be achieved by employing global navigation satellite system (GNSS) which offers a worldwide service coverage with good accuracy. However, installing a GNSS receiver on each device in a network with thousands of nodes would be very expensive in addition to energy constraints. Besides, in indoor or obstructed environments (e.g., dense urban areas, forests, and canyons) the functionality of GNSS is limited to non-existing, and alternative methods have to be adopted. Many of the existing alternative solutions are centralized, meaning that there is a sink in the network that gathers all information and executes all required computations. This approach quickly becomes cumbersome as the number of nodes in the network grows, creating bottle-necks near the sink and high computational burden. Therefore, more effective approaches are needed. As such, this work presents a survey (from a signal processing perspective) of existing distributed solutions, amalgamating two radio measurements, received signal strength (RSS) and angle of arrival (AOA), which seem to have a promising partnership. The present article illustrates the theory and offers an overview of existing RSS-AOA distributed solutions, as well as their analysis from both localization accuracy and computational complexity points of view. Finally, the article identifies potential directions for future research.

Sign in / Sign up

Export Citation Format

Share Document