Search

As COVID-19 spread worldwide, many major grain-producing countries have adopted measures to restrict their grain exports; food security has aroused great concern from various parties. How to improve grain production has become one of the most important issues facing all countries. However, crop diseases are a difficult problem for many farmers so it is important to master the severity of crop diseases timely and accurately to help staff take further intervention measures to minimize plants being further infected. In this paper, a restructured residual dense network was proposed for tomato leaf disease identification; this hybrid deep learning model combines the advantages of deep residual networks and dense networks, which can reduce the number of training process parameters to improve calculation accuracy as well as enhance the flow of information and gradients. The original RDN model was first used in image super resolution, so we need to restructure the network architecture for classification tasks through adjusted input image features and hyper parameters. Experimental results show that this model can achieve a top-1 average identification accuracy of 95% on the Tomato test dataset in AI Challenger 2018 datasets, which verifies its satisfactory performance. The restructured residual dense network model can obtain significant improvements over most of the state-of-the-art models in crop leaf identification, as well as requiring less computation to achieve high performance.

As the COVID-19 spread worldwide, countries around the world are actively taking measures to fight against the epidemic. To prevent the spread of it, a high sensitivity and efficient method for COVID-19 detection is necessary. By analyzing the COVID-19 chest X-ray images, a combination method of image regrouping and ResNet-SVM was proposed in this study. The lung region was segmented from the original chest X-ray images and divided into small pieces, and then the small pieces of lung region were regrouped into a regular image randomly. Furthermore the regrouped images were fed into the deep residual encoder block for feature extraction. Finally the extracted features were as input into support vector machine for recognition. The visual attention was introduced in the novel method, which paid more attention to the features of COVID-19 without the interference of shapes, rib and other related noises. The experimental results showed that the proposed method achieved 93% accuracy without large number of training data, outperformed the existing COVID-19 detection models.

In practice, early detection of disease is of high importance to practical value, corresponding measures can be taken at the early stage of plant disease. However, in the early stage of disease or when a rare disease occurs, there are limited training samples in practice, which makes machine learning especially deep learning models hardly work well while the stronger representative ability needs large-scale training data. To solve this problem, a fine grained-GAN based grape leaf spot identification method was proposed for local spot area image data augmentation to the generated local spot area images which were added and fed them into deep learning models for training to further strengthen the generalization ability of the classification models, which can effectively improve the accuracy and robustness of the prediction. Including 500 early-stage grape leaf spot images every category were fed into the proposed fine grained-GAN for local spot area data augmentation, 1000 local spot area sub images every category were generated in this study. The improved faster R-CNN was integrated in fine grained-GAN as local spot area detector. After that, the segmented sub-images and generated sub-images were mixed as training data input into the deep learning models, while the original segmented sub-images were used for testing. Experimental results showed that the proposed method had achieved higher identification accuracy on five state-of-art deep learning models; especially ResNet-50 had got 96.27% accuracy, which obtained significant improvements than other data augmentation methods and verified its satisfactory performance. This is of great practical significance for rare diseases or limited training samples.

In this paper, an improved discrete sliding mode control (SMC) is investigated for a hypersonic vehicle (HV) cruise control system under packet dropouts. Packet dropouts are common faults in networked control systems (NCS) of HV. The controller designed consist of a discrete SMC controller and a state feedback compensator. The state feedback compensator is designed based on the HV asynchronous dynamic model under Lyapunov theory to restrain the influence of the data packet dropouts, improve the tracking performance and anti-fault robustness. Set the HV system with state feedback as an augmentation system. A discrete SMC system is established for the augmentation system to achieve the target trajectory. We verify the discrete SMC approach in the cruising flight. A simulation example supports the theory.

In this study, a novel exponential artificial potential field is applied to the three-dimensional trajectory planning of high-speed rotor drones. It perfectly avoids obstacles and accurately reaches the target position, and is difficult to fall into the local minimum. However, the selection of gravitational gain coefficient and repulsive gain coefficient in the artificial potential field method is a very difficult problem. In this paper, we choose to use the beetle antennae search optimization algorithm to filter out the most suitable gain coefficient. The BAS optimization algorithm is used for the first time to optimize the gain coefficient of the artificial potential field. Due to its individual search mechanism and update strategy, the optimization algorithm accelerates the speed of iterative convergence and reduces the possibility of falling into the local optimal solution of the algorithm. In the simulation experiment, it can converge to the optimal value after the number of iterations less than ten times. The results well prove that the algorithm can realize the global path planning of the high-speed rotor drones. Moreover, the optimized trajectory planning method has a fast convergence speed and a stable convergence effect, and is suitable for the three-dimensional real-time trajectory planning of high-speed rotor drones.

With the increased adoption of IoT devices, it becomes necessary to detect malware before sensitive information is leaked to untrusted individuals or technologies. Because of their limited computing resources, conventional malware detection techniques cannot be adopted to IoT devices. To solve this problem, we propose a malware detection and prevention framework that tracks information flow within the IoT ecosystem. The framework focuses on Android malware that compromises IoT user information in mobile devices as well as in the cloud.

In order to solve the problem that manipulators can only move in a fixed trajectory in common industrial production at present, we propose a monocular visual positioning method based on artificial beacons, combining visual position technology with the movement of manipulators to complete the reliable path planning of manipulators. When it comes to practicability, the traditional point pair-matching method is equipped with robust factors, to increase the reliability. We adopt a linear predict algorithm and ROI (region of interest) in the solving of PnP (Perspective-n-Point) to ensure the real-time performance and improve the robustness. Finally, the reliability is verified by experiments, which provides theoretical basis and practical reference for the path planning of manipulator based on artificial beacons.

In order to solve the problem of insufficient aerodynamic moment caused by thin air in the re-entry stage of hypersonic vehicle, this paper establishes an attitude angle model of hypersonic vehicle with reaction control system (RCS), and derives its affine linear model to decoupled the internal and external loop. According to the dead zone and saturation characteristics of RCS thruster, a control method to convert continuous moment into discrete switching instruction using pulse width modulation (PWM) is proposed. Since the number of thrusters is usually redundant, the installation matrix of thrusters in the body coordinate is established, and the command moment is coordinately distributed to the individual thrusters. Then a double-loop sliding mode controller (DSMC) is designed to achieve attitude stability and trajectory tracking. Finally, the simulation results show that DSMC has higher maneuverability, fewer thruster switches and stronger robustness to interference.

A path planning method based on Biogeography-Based optimized (BBO) artificial potential field algorithm is presented. The parameters of artificial attraction potential field and artificial repulsion potential field is optimized through state feedback and cyclic iteration. The BBO algorithm is used for parameters optimizing. While involving the optimization, the complexity of the design of path planning algorithm is reduced. The simulation results under local minimum situation and target unreachable situation is done. The results illustrate that the BBO based path planning algorithm is able to effectively accomplish the 2D path planning.

Sensor faults diagnosis is an important research field of the Hypersonic Vehicle (HV). A fast and effective fault diagnosis method can reduce the probability of flight accidents and improve the reliability of vehicle. Aiming at the uncertainty of the HV model and the rapidity of fault diagnosis, wavelet packet decomposition integrated with Biogeography Based Optimization Kernel Extreme Learning Machine (BBO-KELM) is proposed for the first time to identify fault. And the Biogeography Based Optimization (BBO) is used to improve the performance of the KELM firstly. The time-frequency analysis capability of wavelet packet decomposition and the high learning speed of KELM enable fast and accurate sensor faults diagnosis for the HV.

The paper addresses then event-triggered consensus problem for the leader-following multi-agent system with unknown nonlinear inherent dynamics and subject to external disturbances. Based on the neighbors’ sampled information at event-triggering time instants, a distributed control protocol is proposed to ensure that the nonlinear multi-agent system can reach consensus with the desired disturbance attenuation level. Some conditions are further derived to determine the feedback matrix of the controller for networks with directed and undirected interaction topology, respectively. Finally, a simulation is given to verify the obtained theoretical results.

Most of the progresses achieved in radar high range resolution profile (HRRP) recognition rely on the closed set condition, where the test sample is from a known class. In realistic scenario, however, the test sample may be drawn from unknown classes, which is regarded as an open set recognition task. In such cases, conventional recognition algorithms will inevitably make a wrong prediction. In this paper, the open set problem is addressed by incorporating the extreme value theory (EVT) into the random forest (RF) classifier. The outputs of RF are analyzed to determine whether the test sample should be rejected as an unknown class. At the training phase, a Weibull-based extreme-value meta-recognition is introduced to describe the statistical characteristics of the known classes. At the testing phase, a probability estimation method is introduced to compute the probabilities of the test sample belonging to known and unknown classes based on trained Weibull distributions. The test sample is assigned to the class of highest probability. Experimental results demonstrate that the proposed method outperforms the state-of-art NN, 1-vs-set machine and W-SVM in rejecting unknown classes.

With the arrival of the age of Big Data, geosciences data have been increasing at a tremendous speed. The complexity of geosciences data is increasing, and the demand for geosciences data is also expanding. Obtaining target data and associated data from massive multi-disciplinary, multi-sources, and multi-type geosciences data quickly and accurately, and then recommend it to the users is essential to promote the current development of geosciences data sharing. Semantic-oriented metadata association is an applicable approach to solve the association and recommendation of geosciences data. Based on the systematic analysis on metadata of geosciences data, this study presents a semantic similarity calculation model, which is used to calculate semantic similarity for geosciences data attributes. First, a detailed analysis is conducted on several existing geosciences data metadata standards within different research areas, and then metadata elements, which can stand for the geosciences data attributes, are extracted from them. Eight metadata elements between the extracted metadata elements, title, keywords, abstract, data type, data format, discipline, spatial range and temporal range, are selected as associated items. Thereafter, the semantic similarity of each associated item is calculated by taking four kinds of relations, the text relation, the category hierarchy, the time relation, and the spatial relation into account. In addition, to synthesize the results of the semantic similarity calculation of all associated items, the strategy of aggregation of multi-dimensional similarity scores is implemented to get the overall geosciences data semantic similarity. This study investigates the semantic association between geosciences data. The geosciences data can be recommended to the users in order of overall geosciences data semantic similarity. As a case study, the proposed association model has been applied to the retrieval of metadata from the Scientific Data Submission and Sharing Service platform for National Special Program on Basic Science and Technology Research of China. Compared with the retrieval method only relying on keyword matching, by taking consideration of the multiple geosciences data attributes, this model has the advantages that can explore the deeper semantic association between geosciences data, which is conducive to enrich the retrieval results, and has strong measurability, which is conducive to efficiently fulfill the users' requirements. Results of the experiment shows that this model is able to establish semantic relevancy on geosciences data attributes and recommend geosciences data from different research areas to the users. The association method of this study has a certain reference to the development of the accurate discovery, intelligent recommendation and sharing for massive geosciences data.

In this paper, the fixed-time formation control is studied for multi-agent systems with first-integrator dynamics. Using the neighboring state information, the fixed-time formation protocols are designed for both time-invariant and time-varying formation cases. Then for the two cases, it can be demonstrated that the multi-agent systems can realize the expected formation within the prescribed settling time that is independent of initial conditions. To make the system move as a whole after realizing the expected formation, a reference speed input is added into the distributed protocols. Simulations are included to demonstrate the correctness of our theoretical results.

We studied how stego-image is affected by color, image size, and the number of bits used to encode data in terms of delta E, SSIM, PSNR. The study shows that the black color is better to hide data than other colors such as red, green, blue, yellow, fuchsia, cyan, and white. The white color is the least one to hide data. One bit encoding that holds smallest amount data has a higher PSNR and delta E than that of multiple bits to hide data. The black color is sensitive to luminance and responses to contrast change more than other six colors as well.

This paper solves the event-triggered consensus tracking problem for multiple agents with unknown nonlinear inherent dynamics, in the presence of a given mobile leader. For each agent, a consensus protocol is proposed by using the local sampled information at its own event-triggering time instants that are determined by a distributed event-triggered condition. It is shown that all the nonlinear agents can reach consensus under the designed event-triggered protocol, and the calculation methods of the feedback matrix are given for directed and undirected interaction topology, respectively. Finally, a simulation is given to verify the obtained results.

An Intelligent semi-physical simulation system of flight control system of the Hybrid Flight Vehicle(HFV) is proposed. It is designed according to the requirement of flight control system test, demonstration of flight control algorithm, as well as simulation of flight under multiple flight conditions. The hardware of the system is consisted of flight control computer, portable test box, and the simulation computer. The system, portable and universal, setting serial port communications, switch communications, optical communications, data acquisition and flight control in one. Using Real-Time eXtension(RTX) and Microsoft Foundation Classes (MFC) as the development environment, the software of system is based on a modular design concept, giving a friendly human machine interface and easily expanding. The simulation platform has been successfully applied to the ground simulation of the flight control system, completing multi-state flight simulation, achieving comprehensive bus information monitoring, data-storage and display in real time. The system gets the condition that may occur during the flight, laying the solid foundation for the flight test.

The goal of information hiding is to securely hide important information within digital documents such as files, images and videos. In this paper, we develop a new method that can efficiently and securely embed a sequence of binary bits into a gray-scale image in two stages. In the first stage, the sequence is divided into a few subsequences of equal length and a set of one-dimensional logistic mappings are used to shuffle the binary bits in the sequence. In the second stage, the shuffled sequence is divided into regions of equal length and the regions are sequentially embedded into the corresponding pixels in the given gray-scale image. The embedding of a region in the shuffled sequence is performed by replacing the least significant bits of the gray value of the corresponding pixel with the bits in the region. Analysis and testing results show that this information hiding algorithm can securely embed information into a gray-scale image without altering the content of image significantly and thus can potentially be used in practice for secure image-based information hiding.

Supercavitating vehicles (SV) are a class of high-speed autonomous underwater vessels. They present a great challenge in designing the guidance law in comparison with the traditional autonomous underwater vehicles. This is due to the fact that their constraints and working environment are much more complex. To tackle the above challenge, an optimal nonlinear midcourse guidance logic is proposed by considering the wake terminal guidance and the remote target attack tasks. The proposed guidance logic is optimized by using an efficient genetic algorithm to obtain its optimal parameters. The results from our simulation case study suggest that the proposed guidance logic can meet both the motion requirements and navigation constraints while effectively cooperating with the wake terminal guidance. Moreover, it has potential in reducing the energy consumption to significantly improve the overall vehicle energy efficiency.

With the fast development of the Internet, the lightweight 3-dimension (3D) geographic scenes visualization system is expected to play more important roles in web-based GIS. Finding an effective solution to render 3D geospatial data has great significance in Cartography and Geographic Information Systems. The emergence of HTML5, as well as WebGL (Web Graphics Library), provides a new approach for 3D WebGIS. WebGL is a JavaScript API for rendering 3D graphics within compatible web browsers. Other plugins are no longer required for displaying 3D objects. WebGL is regarded as a part of new generation web standards and is completely integrated into most popular browsers. It allows GPU accelerated usage of physics and image processing. WebGL elements can be mixed with other HTML elements and composited with other parts of the page or page background. This paper discusses the design and implementation of a digital city roaming system based on WebGL technology. Users can make cartographic analysis, destination query, data query and map mark with the system in browsers. All 3D models in the scene are converted to glTF (GL Transmission Format), and can be used to load 3D scenes and models. For the complex scenes, this system only downloads the scene data within the current view using visibility culling techniques, then joins these data into the rendering queues. At the same time, the data request task is managed in queue by the management layer using prefetching strategy. On the client-side, a method is proposed to deal with occlusion culling based on the parameters of current scene, geometry shape and positional relation. The client-side removes the occluded model data from memory, while the remaining visible data is passed to the GPU rendering pipeline. Then the client-side renders the scene in browser according to the parameters of these glTF files. The experiment result shows that this system can smoothly render the city 3D scene in the browser without any plugins. The city roaming system based on WebGL shows a good interactive experience of 3D visualization for users. The paper provides a very effective and promising method for establishing 3D WebGIS realization.

In this paper, we use SLR normal point observations provided by EDC to validate AIUB zero-difference kinematic orbit of GRACE satellites. By selecting 4 representative SLR sites, they are analyzed in detail from the regularity of observed elevation angle, azimuth angle and residual value of GRACE-A/B satellites. The results show that the data quality of different SLR sites is uneven, and more than 97% of the SLR observed data are very stable. There is almost no systematic error in the SLR validation of GRACE kinematic orbits, and the accuracy is better than 2.5 cm for GRACE-A and 2.7 cm for GRACE-B respectively. Under the statistics of nearly 350,000 available observations for each satellite, the accuracy indicators have strong similarity in the same site. The SLR residuals of each site obey normal distribution. The number of SLR data is inversely proportional to the observed elevation angle, and the geometric distribution of spatial observation points has a periodicity of π at different azimuth angles.