It contains plenty of ocean wave and sea surface current information in the sea clutter images formed by X-band marine radar’s echo. Applying the method to calculate the significant wave height from the SAR imagery, which supposes the significant wave height in linear relation with the square of the signal-to-noise ratio of radar images, the significant wave height has been obtained from estimating the images of X-band radar. The experimental data were analyzed in the Small Mai-island sea area. Firstly comparing the effect of filtering direct current versus estimating result, deriving the significant wave height estimated by counting the signal-to-noise ratio after filtering direct current which is match better; then according to wave height measured by wave buoy, analyzing low and high wave height to do linear fit and gain calibration coefficient separately, the significant wave height evaluated is all the more precise.
Precipitation, as one of important input variable for land surface hydrologic and ecological models, shows a high spatial and temporal variation. The daily precipitation by manual measurement is collected in Koxkar glacier catchment in rainy season during the period of 2009–2016, and hourly precipitation by standard tipping rain gauge is collected from May 19 to August 22 in 2016. Comparing the manual and standard tipping rain gauge, the performance of standard tipping rain gauge is suited in Koxkar glacier catchment in rainy season. The distribution of hourly precipitation is analyzed. The results have shown that precipitation is occurred more in daytime than in night, and frequency is concentrated at 15:00 to 17:00. The trend of precipitation and precipitation days are consistent at night, but obvious inconsistency in the daytime. These characteristics of precipitation have important role for further understanding of glacier changes with climate change.
The recent advancement in wireless communication has motivated increasing number of mobile applications, including computing-intensive tasks. However, it takes resource-limited mobile devices a lot of energy to execute these tasks. Computing offloading is helpful in the scenario, where mobile device offloads part of the task to available devices. In this paper, we propose an algorithm AOA (Alternately Optimizing Algorithm) to alternatively optimize task and power allocation in order to achieve the minimum system energy consumption under given time constraint. KM (Kuhn-Munkres) algorithm in graph theory is adopted to get the optimal task assignment. And we get the optimal solution for power allocation via mathematical derivation. Simulations have shown that the proposed algorithm can give a global optimal task and power allocation solution.
This article proposes a stochastic multilevel-state framework to model the animal’s behavior. The motivation of this article is the variety behavior influenced by several factors, which can lead to the state explosion. The proposed framework processes data from an automated sensing system and constructs the model. The data gathered from captive Antelope goral (Naemorhedusgriseus) in Chiang Mai Night Safari, Thailand. The data is gathered from the activity and the environmental factors in the cage by none invasive method. The model separated observed data into two main classes: the upper-level data and the lower-level data. The upper-level data represents the environment data such as temperature, humidity, and light density. Moreover, the landscape of the captivity area also takes into consideration. On the other hand, the lower-level represents the location of the animal of interested in the captivity area. The working strategy of this work is to cluster the each type of data and link them together by a stochastic approach. Both layers of data will be handled independently in clustering algorithm and determine probabilistic of state transaction. From the data observed from the sensor, the Probabilistic Automaton (PA) function is constructed. It is a function of producing the next stage based on the previous behavior states. The initial state of the framework is in the lower-level data (the current location of the animal). Then, the PA using the current lower stage and the current upper stage generates the next stage, location of the animal. Both the lower stage and the upper stage are traverse along the constructed automaton. The result can suggest computing methods, which can utilize to zoo research, which performs behavior monitoring, and in the other studies area, or subject of study, such as, air pollution dispersion that tracks the movement of pollutant according to the environment. The benefits of the proposed methods also can be used to create the application to attract the tourist to the area, which animally is likely to display themselves.
Automatically detecting ground object from optical remote sensing images has attracted significant attention due to its importance in both military and civilian fields. However, the diversity of configuration for different object and the complex background information makes this task difficult. Moreover, the high-level semantic information is usually ignored. To address these problems, we propose an efficient method that extracts deep feature with high-level semantic information from a classification convolutional neural network, and separates the regions of interested based on deep feature. Then each region of interest will be sent to another convolutional neural network to verify whether they are true objects or not. Our proposed method can adapt different objects. Also, it doesn’t need any bounding box information for training. We build two remote sensing datasets, SROD-3 and RSHOA-4, to evaluate our detection method. Experiment result indicates that our detection method performs better than other state of the art methods, including Faster-RCNN and YOLO9000.
Natural disasters frequently occur all over the world in recent years. Current researches show that a disaster often causes different kinds of secondary disasters. A good understanding of the chain reaction in disasters can provide guidance for disaster prevention and mitigation. Most of current researches analyze the disaster from the perspective of the disaster mechanism such as the geo-statistical model. This paper proposed an intelligent method of discovering the relationship of disasters using big scholar datasets. This method does not investigate the mechanism of disasters themselves, but analyze the relationship among disasters from the perspective of big data mining. The experiment results show that it is able to get reasonable relationship of disasters without much human interventions. The proposed method will enlighten many other knowledge-discovering applications in geospatial domain.
In this paper, a deep learning classification model is proposed for automatically detecting the marine oil spill in Lanset-7 and Lanset-8 images, which can combine fully convolutional network (FCN) with Resnet and Googlenet respectively. The classification algorithms, i.e. FCN-Googlenet and FCN-ResNet are compared to the state-of-the-art Support Vector Machine (SVM) method. The experimental results show that our FCN-Googlenet and FCN-ResNet models outperform other approaches with a significant improvement. Moreover, our methods are more flexible in that no restriction on the size of input image is required in our algorithmic setups, which is more suitable in real applications.
Keyword search in relational databases has been widely studied in recent years. Most of the previous studies focus on how to answer an instant keyword query. In this paper, we focus on how to find the top-k answers in relational databases for continuous keyword queries efficiently. As answering a keyword query involves a large number of join operations between relations, reevaluating the keyword query when the database is updated is rather expensive. We propose a method to compute a range for the future relevance score of query answers. For each keyword query, our method computes a state of the query evaluation process, which only contains a small amount of data and can be used to maintain top-k answers when the database is continually growing. The experimental results show that our method can be used to solve the problem of responding to continuous keyword searches for a relational database that is updated frequently.
The recommendation quality of new users plays an increasingly important role in recommender systems. Collaborative Filtering cannot handle the cold-start problem, while the content-based approach sometimes can achieve recommendation with new items. To recommend in the wallpaper field, this paper proposes a content-based recommender system and extracts the features of wallpaper via the deep learning approach. The first part of the recommendation model is the convolution layers, and the model takes the output of full connection layer as features to employ. In order to improve the scalability, the model adopts deep neural network as non-linear dimension reduction method to reduce the image features. Taking the recommended results into account, this paper compares the feature similarities of user images and those in the image library. Finally, the model sorts them via cosine similarity, and presents the recommendation results using Top-K list. In the experiment, our model is trained with selected wallpapers on MIRFLICKR dataset, and uses VGG on ImageNet for feature extraction. The experimental results indicate that WPNet will have higher hit rates with different K if the image division of some wallpapers can be improved, and achieve a better performance in less time under the recommendations of new items.
The semiarid mountainous region is characterized by sparse vegetation and rich source of loose deposits, which is favorable for the formation of debris flows. Benzilan-Changbo segment in the upper reaches of Jinsha River is selected in this study. Based on field investigation and interpretation of remote sensing images, the development characteristics of regional debris flows in the semiarid area are studied. Six assessment indices are selected, including lithology, structural fault, slope gradient, relative height of the watershed, annual average rainfall and normalized vegetation index. Based on GIS, the assessment model of debris flow susceptibility for semiarid region was built using AHP (analytical hierarchy process) method, so as to obtain the grid map of debris-flow susceptibility zoning in the study area. The study area is divided into small watershed as one unit for regional statistical analysis and classification. Finally, the debris-flow susceptibility assessment map based on small watershed analysis can be obtained. The assessment results show that the very high susceptibility area and high susceptibility area of debris flow are mainly distributed in the northeast, central and southwest banks of Jinsha River, with an area of about 1040.9 km2, accounting for 35.7
Building roof temperatures mainly affected by solar radiation. With the solar radiation intensity changing, the change of roof temperature also occurs constantly. It has an uncertainty to a high degree, so the grey system can be combined with data analysis for researches. Based on the classical NDGM model, this paper introduced the fractional order NDGM^qp model to study the important properties of the model and used the PSO particle swarm optimization algorithm to optimize the fractional order. Finally, the two representative measuring points, namely, the maximum solar radiation and the second solar radiation points, were taken as the experimental objects. The experimental results show that the mean absolute percentage error (MAPE) of the optimized fractional order NDGM^qp model for the roof temperature is several percentage points higher than that of the classical GM, DGM and NDGM models, as well as the minimum error of the model can reach 2.4247
Based on the approximation that tracklet kinematic association likelihoods satisfy the Markov or path-independence assumption, several polynomial-time bipartite matching algorithms were proposed to stitch track segments for their effectiveness. However, with target density increasing, their stitching performance would degrade inevitably. Despite the help of feature information, it is remarkable that the aforementioned approximation is no longer valid since the feature information is usually sporadic. In order to solve this problem, track graph is utilized and the feature information is passed through the graph to calculate the tracklet feature association likelihood under path-dependence assumption. It makes bipartite matching algorithms valid again. Finally, simulation results demonstrate that the proposed algorithm outperforms previous algorithms based on path-independence assumption in the dense target situation.
In this work, we present the design of sensor platform for air pollution monitoring. During the design process, we took into account a lot of problems such as system architecture, power consumption and linearity consideration. ADC plays a vital important role in high-linearity sensor micro system, several practical techniques which can improve the performance and decrease power consumption of ADC are discussed in this paper.
In this paper, the physical layer security in decode-and-forward wireless full-duplex relay network was investigated. In this scenario, an eavesdropper was present. For improving the physical layer security, the source and relay not only transmitted message signal but also transmitted jamming signal to interference with the untrusted eavesdropper, which needed not external friendly jammers. How to allocate the power to transmit message signal and jamming signal was a problem. Furthermore, the constrained optimization problem was formulated and the optimal power allocation solution was derived. Compared with other cooperative jamming schemes, our proposed scheme will effectively improve physical layer security of the legitimate user. And simulation results verify the properties.
Digital mining and unmanned mining have become the development trend of the coal industry at present. Aiming at the deficiency of the existing safety monitoring technology in coal mine in the aspects of progressiveness, reliability and real-time, this paper puts forward the key technology of networked coal mine safety production monitoring. The structure of the communication mode based on “one network one station” is discussed in this study, and the design method of the specific network is analyzed. The accurate locating method of mine moving target based on TOF is studied. The technology of mine fire monitoring system based on wireless self-networking is studied. This paper is of theoretical and practical value to improve the development of coal mine safety monitoring technology and the level of mine safety control.
Currently most research of virtual scene simulation focus on the establishment algorithm of three-dimensional terrain model in the scene, but rarely consider the communication mechanism, multi-thread synchronization, dynamic loading issues of the distributed three dimensional virtual scenes. In this paper, the simulation engine based on vega prime is achieved by multi-threading technology, and communication protocol is implemented by winsocket technology, then the network architecture with high cohesion and low coupling control/running terminals are built. Combined with dynamic loading model reuse technology, three-dimensional virtual scene system based on distributed network communication is achieved. Based on the algorithm interface reserved in the system, the multitasking cooperative swarm intelligent pathfinding algorithms can be integrated. Then the communication model, multi-thread synchronization, dynamic loading and pathfinding problems are solved in the distributed multitasks three-dimensional visual scene system. The simulation results show that network data communication and the routing algorithm simulation are realized in three-dimensional virtual environment, and the effectiveness and feasibility of the algorithm can be verified, thereby the cost and risk of late operation are reduced.
In order to solve the problem of frequent spectral state transition in the traditional cognitive radio network, the existing spectrum sensing is less reliable and the “hidden terminal” is added to reduce the interference to the main user. In this article raised introducing multiple secondary users to cognitive radio network and carrying out Hidden Markova Model (HMM) to main user’s spectrum. Recursion calculating forecast probability of user’s next time slot spectrum status is “busy” or “leisure.” All counting of the secondary users “busy” and “leisure” frequency. If the percentage is “busy” exceeds a certain value, so could judge following time slot spectrum status is “busy,” otherwise it is “leisure.” The simulation results show that the algorithm is 10%–20% higher than the average energy sensing algorithm, and it is more obvious at low level. This paper improves the perceived reliability while rapidly detecting the spectrum, and greatly reduces the interference to the primary user.
The Scale Invariant Feature Transform, SIFT, has good ability to detect very stable feature points. But at present, there are very little researches on SIFT in our country, and most of them are concentrated in the areas of Image Registration and Image Stitching. In this paper, SIFT and KLT will be combined for feature points detection and tracking. First, SIFT algorithms is used to detect stable feature points, and then the KLT method is used to track the feature points. The experimental results show that the new method provides a good method in the field of feature points detection and tracking.
In the light of the fact that the detection data is small, the fault diagnosis of the wind turbine often occurs. A multi-channel data acquisition based on wireless transmission wind turbine fault monitoring device is presented in this paper. The system used sensors to collect fan status data, and processes the signals by 32 bit digital signal processor DSP (TMS320F28335). The LabVIEW software system and the MCGS (Monitor and Control Generated System) configuration software display synchronously. The simulation experiments show that compared with the traditional wind generator online detection device, it can collect more data under various operating conditions, and provide a reliable basis for fault diagnosis.
Localization is one of the key issues of wireless sensor networks. Because of the energy and hardware constraints of sensor nodes, we usually use RSSI (Received Signal Strength Indicator) as a ranging method. In this paper, we proposed an RSSI-based localization algorithm, which takes use of the RSSI values received by sensor node from mobile anchor node to estimate the position of sensor node. We used mobile anchor moving along specific trajectory to locate the unknown nodes, study four different trajectories and analyze the simulation result. Our research indicates that reducing the time interval of transmitting beacons can improve the positional accuracy when using as few anchor nodes as possible. The relative position of anchor’s trajectory and the unknown node has an influence on the location result, and an appropriate trajectory can optimize the localization accuracy.
