Abstract:In this paper, we propose a model called IRMatch for matching images and sentences based on implication relation to solve the nonequivalent semantics matching problem between images and sentences. The IRMatch model first maps images and sentences to a common semantic space respectively by using convolutional neural networks, and then mines implication relations between images and sentences with a learning algorithm by introducing maximum soft margin strategies, which strengthens the proximity of locations of related images and sentences in the common semantic space and improves the reasonability of matching scores between images and sentences. Based on the IRMatch model, we realize approaches of bidirectional image and sentence retrieval, and compare them with approaches using existing models for matching images and sentences on datasets Flickr8k, Flickr30k and Microsoft COCO. Experimental results show that our retrieval approaches perform better in terms of R@1, R@5, R@10 and Med r on the three datasets.

Abstract:This paper summarizes the literature about the following aspects: the definitions of customer churn and customer churn management; research contents and application scenarios of customer churn issues; customer churn prediction algorithms and feature extraction methods; the evaluation technologies and measurements. In the end, we point out the shortcomings of the current research and put forward some future research directions.

Abstract:Information spreads quickly on social networking platform. In order to effectively carry out public opinion early warning and quantitatively evaluate the importance of users in social network, the fuzzy comprehensive evaluation method is introduced into the user influence modeling problem. Based on the analysis of the behavioral analysis of the user's behavior on the social platform, the evaluation system including five indicators, such as user active number of fans and average forwarding number is constructed. A new fuzzy synthesis operator is proposed to construct the user influence evaluation model based on the shortcomings of the traditional fuzzy comprehensive evaluation algorithm in calculating the weight of the evaluation index. This operator can adjust the weight of the impact on the evaluation results according to the demand. Using the real data of Sina microblogging social platform, combined with comparative experiments and practical assessment, the method can more accurately reflect the actual impact of the user in the social network.

Abstract:The emergence of HPC cloud has inspired service provider to deploy GPU in the cloud ecosystem (e.g., Amazon EC2 GPU instance, Aliyun GPU Server). GPU as a computing accelerator is playing an indispensable role in clouding computing. Due to the intrinsic sharing feature of cloud, GPU sharing does not only boost the utilization, lower the cost, but also makes it easier to manage. GPU virtualization comes to solve this problem through Hypervisor and cooperation of software and hardware. This paper collects the methodologies of GPU virtualization and makes a classification and analysis. In addition, it concludes the existing problems and proposes the future works of GPU virtualization.

Abstract:The motor speed accuracy is an important performance index of the servo system. The traditional PI control cannot meet the high precision requirements of servo system. The application of advanced control strategy to permanent magnet synchronous motor is the development trend of servo system. This paper introduces the application of the precision velocity measurement method and the advanced control strategy of the permanent magnet synchronous motor in the motor speed system, analyzes the advantages and disadvantages and the scope of the various methods, summarizes the latest research results and the problems to be solved, and the future The research direction is forecasted.

Abstract:The face recognition technology is widely used for its low cost, user-friendly and high efficiency. At the same time, identity forgery attack has also been the corresponding occurrence. The face recognition system attacks include photo face attacks, video face attacks and three-dimensional face model attacks, etc. For these attacks, prevention methods are carried out around the midpoint of in liveness detection based on human face. This paper focuses on the blink detection and background analysis algorithm, and carries out eye location with regional growth algorithm. The morphological operation is used to judge the human eye state, and the Hash algorithm is used to compose the background difference. These methods construct a Liveness Detection Systems. Based on the blink detection and background analysis algorithm, this paper designs a liveness detection system including blink detection module and background analysis module; uses the MFC architecture and OpenCV2.4.9 to build a liveness detection system which can resist photo attack and video attack; makes the experiment and evaluation of the system. In comparison with other similar types of systems, the system performance of this paper is excellent.

Abstract:In this paper, aimed at the complicated characteristics, patterns and rules of the future “let go” electricity market, smart contracts of electricity transaction based on the blockchain platform are proposed. Then the key technological difficulties are analyzed and solutions are given. By running a smart contract instance in a peer-to-peer network composed of 4000 nodes, the success rate is 99.38% and the average time consumption for each transaction is proven to be 16s. If our method is applied, we can in the field of electric power market transaction settlement through different systems, reduce the trust of the electric power market transaction cost, reduce the friction in the process of liquidation, improve the efficiency of the clearing and settlement.

Abstract:The video surveillance technology has a wide application prospect in traffic management, public safety, intelligent city, and is developing towards intelligent recognition, real-time processing, and large data analysis. In this paper, we propose a new system for large-scale real-time video surveillance. The system is based on Spark streaming, distributed storage and OLAP framework so that multi-channel video processing has obvious advantages in scalability, fault tolerance and data analysis of the multi-dimensional polymer. According to video processing algorithm, the processing module is divided into single machine processing and distributed processing. The video processing is separated from the data analysis, and the further operation of the multi-channel video output data is completed by using Kafka message queue and Spark streaming. Combining the distributed storage technology with OLAP framework, the system achieves real-time multi-dimensional data analysis and high-performance real-time query.

Abstract:The virtualization technology research is gradually moving from high-performance server to mobile intelligent devices. Existing virtualization solutions are mostly multi-core solution with high system overhead and low efficiency. This paper presents a lightweight virtualization scheme based on Linux container for multi-screen display and resource limitation of vehicle system. Our program, through the use of Namespace resource isolation mechanism and Cgroup resource control mechanism, can start several Android virtual machines on the ARM platform at the same time, while displaying on different screens and running independently. The performance test results show that on the ARM platform, the program uses less than 7% of the memory of the two systems, and the average CPU usage after virtualization is only 1% higher than the native Android system.

Abstract:Target recognition has been the hot issue in the field of artificial intelligence. In order to enhance the efficiency of target recognition, this paper proposes a method based on multilayer feature extraction of convolutional neural network. By inputting images into convolutional neural network for training, this method implements feature extraction at each full connection layer of network, inputs the features obtained into classifier, and then compares the output results. The lower full connection layer activated by relu function is selected as feature extraction layer, whose recognition rate is higher than that in higher full connection layer. This paper builds up office supplies dataset, and realizes the office supplies identification system based on the multilayer feature extraction of convolutional neural network. The layer relu6 of AlexNet is selected feature extraction layer, and the optimal training image quantity as well as the optimal classifier construction system is chosen, which verifies the feasibility of this method.

Abstract:Meteorological data interprovincial sharing service provides the weather prediction and meteorological service of timely data for neighboring provinces. The National Meteorological Information Center now gets statistics and publishes the transmission quality of interprovincial sharing service by month. It is not easy to find out the abnormality during data transmission timely. We design and implement a real-time monitoring system for interprovincial data sharing service using the monitoring method for data upload transmission. Firstly, the system above decodes AWS-data and then shows the receiving and sending status of the hourly national and regional automatic station data and radar data. This system provides data application department and management department of the daily, monthly and annual statistics results based on SQL Server database and it also generates interprovincial sharing log files. The application result shows that the real-time monitoring system of meteorological data interprovincial sharing service is stable and reliable.

Abstract:With the booming of performing arts in China, which is obviously embodied in the increasing number of performances and audiences, and the requirement of performance quality is increasing. Therefore, a virtual choreography system becomes particularly important with great value and significance. This paper aims to explore and develop a virtual choreography system for medium-scale and large-scale performances. By doing research on the key technology of the virtual choreography system based on WebGL, we put forward a series of methods including file compression, preloading, embedded server and v8 engine optimization, which can help assure the smooth operation of the system. Meanwhile, we adopt WebGL to achieve the creation and rendering of three-dimensional models so that users can directly complete the whole process of large-scale rehearsals in the system. It has great prospect.

Abstract:In order to meet the multi-directional and multi-level testing requirements for the WAMS advanced application system, and to meet requirements of the test system on flexibility, reusability and scalability, a universal test platform is designed and constructed. In this paper, the basic function, composition, software structure, data model, application test cases of the universal test platform are described in detail. The division to the function of the test platform according to structured, hierarchical, modular standard is completed. To provide communication interface with external program and a complete test case according to different input data source cases, the test platform scalability and versatility is realized. The realization of WAMS advanced application system test using the universal test platform significantly reduces the workload of testing, it can also reduce the development cost of WAMS advanced application system and speed up the development progress.

Abstract:In order to realize the goal that the civil unmanned cargo aircraft (UCA) would be controlled by the operator in the ground control station (GCS), the design scheme of the human-machine interface (HMI) for GCS is proposed which integrates the flight instrument and maps. By the way of mixed programming using the tools of GL Studio and VC ++ 6.0, this paper achieves the design of the flight instrument display interface. With the use of VC ++ 6.0 on the secondary development of Google Earth, this paper achieves the design of map display. With UDP transmission protocol, the flight simulation data from the Microsoft simulation flight X is sent to the interface of the display. The test results show that the system can provide complete information for the operator of the ground control station, improve the decision-making efficiency of the operator, and provide a new method for studying the interface efficiency of the ground control station.

Abstract:Big data Privacy security is becoming the hot spot in the various social industries, because attackers can build an integrate portrait to threaten privacy of users by identifying accounts in different sites. Simulation assessment of the attacker re-identification ability is the precondition of users' privacy protection. Therefore, this paper proposes a high similarity algorithm in same day with same behaviors. The core idea of the algorithm is as follows: if a couple account issues similar or identical content on the same day, which also appears many times in different websites, then these two accounts may belong to a person with a high possibility. In addition, this paper builds a new weighting model for the users' attributes to improve the accuracy of user re-identification. After the experiment on more than ten thousand users of the two major domestic social networking site, this algorithm proves to be effective. Experimental results show that even if attacker don't consider users' social relations, the users' tweets, attributes, still provide enough information to make the attacker correlate their different accounts, which will lead to leak of more privacy.

Abstract:The automatic calibration of hydraulic model aims to improve the accuracy of the model of water supply network intelligent management. Currently, the genetic algorithm is widely used for automatic of hydraulic model. In view of problems that the standard genetic algorithm has slow convergence and can be easily trapped in local optimal, the paper makes some improvements of this algorithm. Simulated annealing algorithm is used to stretch the fitness function and the roulette wheel selection method combining elitism strategy is replacing the traditional selection method. Besides, the similarity function is added to avoid the breeding with closest relatives in cross operator and the double convergence criteria is used to reduce unnecessary computation time. The improved genetic algorithm is used to calibrate the water supply hydraulic model of G. The results show that the improved genetic algorithm has better efficiency and accuracy.

Abstract:Nowadays, the widespread phenomenon of flight delays does not only increase massive flight costs, but also impacts the experience of passengers. Reasonably adjusting the aircraft queue in terminal areas can raise the utilization ratio of the runway and reduce flight delays. Finally, the cost of delay can be cut down. To resolve the problem of aircraft scheduling in terminal areas, this article puts forward an genetic algorithm including double crossing operator. Different crossover operations are carried out for chromosomes with different fitness so that the quality of chromosomes can be protected and the others continue to evolve. At the same time, reordering operator is introduced to optimize the descendants after mutation to improve convergence rate of genetic algorithm and makes it more suitable for practical use. The experimental results show that the convergence rate of algorithm is improved and the feasible solution can be obtained within acceptable time.

Abstract:To improve the efficiency of collision detection in complex scenes, a collision detection algorithm based on topological spatial grid is proposed in this paper. Because there are numerous objects with complex shapes, different scales and motion states in the scene, the algorithm firstly partitions the scene with uniform octree, builds oriented bounding box hierarchal tree, sets up the topology of spatial grids and conducts static large-scale object decomposition for accurate positioning in preprocess stage. Then, the algorithm excludes plenty of non-intersect objects with topological spatial grid and projection intersection test, by using bounding volume hierarchy to conduct accurate detection of potential collision pairs and calculates collision points in real-time detection process. Experimental results show that this algorithm improves the real-time detection efficiency, which is suitable for collision detection in complex virtual scenes.

Abstract:Filter functions play a significant role in image processing. The traditional implementation method takes the window as the processing unit, whose size is so small that the pipeline is interrupted frequently. This paper proposes an optimization method of algorithm slicing to settle this problem. First, the filtering algorithm is sliced so that each slice processes one pixel in the filter window. Then, the effective processing range is calculated from the position of the pixel in the filter window to remove the complex conditional judgment of the edge of the image. Finally, the software pipeline is carried out in the column direction, allowing the pipeline to repeat a large number of identical instruction sequences. Combined with BWDSP 1042 special instructions and hardware logic, the median filter and other image filtering functions are optimized. The experimental results show that the performance of the entire image filtering functions is improved by more than 51 times in the four-cluster pipeline mode.

Abstract:During the object tracking, when occlusion occurs, the traditional superpixel tracking algorithm will add the superpixels of the non-target area into the feature space. In the calculation of the candidate sample confidence, the nearest neighbor superpixel in the feature space is used to delimit the cluster attribution of the superpixels in the sample, and the accumulation of the classification error is caused by the excessive number of neighboring superpixels. To solve the problems above, we propose a robust superpixels tracking method. This algorithm uses Bayesian algorithm as the framework. Firstly, we slice the first few frames into superpixels, extract the feature, use the mean shift clustering algorithm and representation model based on superpixel to classify and calculate the class confidence value, and put the feature into feature space. Secondly, the suitable numbers of neighbors can be found with the mean center error of next few frames. Last but not least, during the tracking process, the superpixel is segmented in the specified area of the acquired frame, to extract the feature. The cluster is confirmed with soft classification and the confidence value is calculated. According to the previous frame target position, the Gaussian sampling is collected. We can obtain the sample confidence value with the accumulation of the confidence value. In case of severe occlusion, the sliding window update and the appearance model modification are not carried out, and we continue to use the current model to track. Compared with the traditional tracking algorithm based on nearest superpixel, the algorithm can effectively improve the tracking success rate and reduce the average center errors.

Abstract:Text similarity check is mainly used in Re-check detection of Papers, the deduplication of search engines and other fields. However, it's extremely fussy to extract feature items with the traditional methods for computing the text similarity. In addition, it will bring uncertainty to select elements randomly. To solve these problems, a text similarity method based on improved Jaccard coefficient is proposed. This method takes into account the weights of elements and samples in the document, even the contribution degree to multiple text similarity. The results suggest that the text similarity method based on the improved Jaccard coefficient has been proved to be effective with a satisfactory accuracy, which can be applicable to various lengths of Chinese, English documents. It effectively solves the problem of inexact computing with existing technologies.

Abstract:Due to the constraints of data acquisition time, irradiation dose and geometric position of imaging system scanning, the technology of computer tomography(CT) can only get the data in the limited angle range or the less projection angle at present, which are incomplete angle reconstruction problems. Therefore, the image reconstruction algorithm becomes particularly important. This paper will apply some existing regularization super-resolution reconstruction algorithms to CT images and give a series of comparative analysis, with the effects of reconstruction analyzed under different algorithms. Firstly, the low resolution CT images are registered, and then the spline interpolation is used to enlarge the image. Finally, the image is reconstructed by using the regularization algorithm. Experimental results show that the application of the regularization algorithm can improve the image resolution to a certain extent, and the reconstruction effect is the best with the bilateral regularization, and the L2 norm based total variation regularization is less effective.

Abstract:In order to solve the noise sensitive problem in the traditional Canny edge detection algorithm, based on the curvature estimation of curve, the edge curvature degree operator is constructed according to the noise edge characteristics which are large curvature and small curved space. The edge curvature degree operator is modified with the results of edge detection of the large scale Canny algorithm so that the operator can accurately characterize the distribution of the noise intensity in the two-dimensional space. On this basis, an edge detection algorithm based on edge curvature degree operator is proposed and the edge curvature degree operator can restrain the noise. The experimental results show that the algorithm cannot only suppress the noise effectively, but can also preserve the rich details edges and the slowly changing contour edges, and the detection results show the original structural features of the image.

Abstract:In order to solve the problem of identifying depression tendency among students on sina microblog platform, this paper proposes a depression tendency identification model. By inviting students widely to fill in the self-rating depression scale online on campus we can get the students' score. We collect students' microblog text and ask the psychology teacher to annotate the microblog artificially. In the pretreatment stage, we use the depression emotional dictionary to reassemble the depressed emotion words that are split at the segmentation stage so as to improve the recognition accuracy rate. And then we build a classifier based on the support vector machine to train the data. Through continuous learning and feedback, we get a better classification result. Finally, this paper defines the depression index and uses it to measure the degree of depression for a period of time. The experimental results indicate that the degree of depression measured by depression index is approximately consistent with the results of the scale, the accuracy of the method being 82.35%.

Abstract:The urban traffic system is gradually changing from a single mode to interconnected multi-mode. In order to express the multi-mode traffic network system more accurately and meet the requirements of personal route planning and time prediction, this paper selects Oracle spatial network data model as the modeling foundation. The model takes Wuhan City as an example, and selects the road network, bus network and railway network to construct a multi-mode traffic network. Also the schedule of bus lines and subway lines is added to allow the travelers to find the quickest path to destination and to estimate the time required for the route.

Abstract:Cache is one of the key technologies of named data networking(NDN). However, the basic cache scheme LCE (leave copy everywhere) in NDN leads to much redundancy. The RCOne scheme chooses the cache node randomly without using any information of content and node, which is limited in improving cache performance. The Betw scheme results in that the node has the more frequent replacement with the larger betweenness centrality, which will decrease the cache performance when the node's cache capacity is far smaller than the total content amount. In order to solve those problems, a cache scheme named HotBetw is proposed in this paper based on content popularity and node betweenness to choose appropriate cache node along the content delivery path. The simulation results show that the HotBetw cache scheme can achieve higher cache hit ratio and reduce average request hop compared with existing schemes.

Abstract:This paper analyzes the military equipment logistics center location problem and based on the analysis, builds the hybrid algorithm model which combines fuzzy clustering and genetic algorithm. The core technology is to integrate the fuzzy clustering network model into genetic algorithm in the process of building groups, which can effectively avoid the possibility for genetic algorithm to be prone to premature. The algorithm has been validated to be robust and reliable. The simulation results offer certain references for policy-makers to select the location more scientifically.

Abstract:In recent years, with the acceleration of urbanization process, people become more concerned about security in public places like banks, official buidings and schools etc. Therefore, intelligent monitoring has become a hot topic in current research. This paper mainly studies the face abnormal occlusion events based on the indoor patrol car. Firstly, we extract the foreground of surveillance video. Then, we locate shoulders and use an ellipse to fit the area of head based on the foreground. And then, we determine the face area through the skin color rate. Lastly, we detect the eyes and mouth of the area through Haar detector to determine whether there are abnormal occlusion events. The experimental results show that the algorithm proposed can detect abnormal occlusion events effectively.

Abstract:In order to improve the accuracy of seed selection and shorten the cultivation period of cultivars, the improved random forest algorithm is used to construct the evaluation model of the history data of wheat breeding. Before training the classifiers, the improved SMOTE algorithm is used to improve the non-balance of the training samples. After the training of the base classifiers, we test every classifier's performance and delete bad classifiers to realize the screening of the base classifier in random forest. The experimental results show that the proposed algorithm has achieved good results in wheat germplasm evaluation, which can help to breed varieties.

Abstract:In the traditional AdaBoost algorithm, there are over-fitting problems caused by noise samples. In this paper, an improved AdaBoost algorithm based on noise detection is proposed, called NAdaBoost. According to the traditional AdaBoost algorithm, in the misclassified samples, noise samples vary widely in some attributes. NAdaBoost can, instead, determine the noise samples based on this, and then reuse the algorithm to classify the two types of samples, and ultimately achieve the purpose of improving the accuracy of classification. The experiment on the binary classification shows that the proposed algorithm has a higher classification accuracy compared with the traditional AdaBoost algorithm, as well as relative improvement of algorithms.

Abstract:In view of the problems like the difficulty in memorizing passwords, privacy issues and fake information, a new algorithm combined the static and dynamic features based on the android platform for online handwriting recognition is proposed. The proposed algorithm extracts the static texture and dynamic vector features by adopting the combination of text-dependent and text-independent ways on the new smart mobile devices platform. The problems of handwriting acquisition, fake information and less accuracy are solved. It's safe and fast for identification on mobile devices. The experimental results show the proposed algorithm has a great performance in stability, repeatability, accuracy and safety. It can effectively block the attack from intrusion and is high in security.

Abstract:With the development of seamless connection technology, multimedia applications as the core of the radio technology must meet user networks' requirements for service quality. For heterogeneous wireless networks, switching between different technologies and management domains may occur, and the handoff decisions is no longer based on a parameter but on a multi-attribute synthesis. This paper studies the determination of the weight value of each parameter in multiple attributes. Through the analysis of three kinds of weight assignment methods: analytic hierarchy process, coefficient of variation method and entropy method, the heterogeneous wireless network metrics are evaluated under different business types.

Abstract:With the development of geo spatial information technology, it is more important to build large scale virtual terrain scene with massive spatial data. However, in the face of massive terrain data, how to simplify terrain, improve rendering and rendering efficiency, is the key to the terrain rendering. After the research of the terrain rendering technology of LOD, the analysis and processing of large scale data sets, parallel computing and other related technologies, the parallel rendering technology of massive terrain data based on LOD is proposed. At first, quadtree LOD is used to simplified terrain, secondly it is combined with multi-core CPU parallel computing method to enhance efficiency, then it is combined with the large data scheduling strategy, finally it realizes the parallel rendering of massive terrain data, and it analyzes the non-parallel and parallel experiments under the same conditions. The theoretical and technical achievements in the research can provide a new idea for large scale scene rendering.

Abstract:A large number of research behavioral methods are focused on detecting simple actions such as walking, jogging, or jumping, while less research is on violence or aggressive behavior, but these studies are useful in some surveillance scenarios, such as: Prison, self-help banks, shopping malls and so on. Traditional methods of violent behavior recognition research mainly use a priori knowledge to manually design features. In this paper a violence detection method based on 3D-CNN structure is proposed. The three-dimensional deep neural network directly manipulates on the input, which can be a good extraction of violent behavior of time and space characteristics of information. It can be seen from the experimental results that this method can identify the violent behavior better than the characteristics of hand-craft features.

Abstract:In this paper, the cutting force and vibration signals in different axial directions and the RMS of the acoustic emission signal in the milling of the high speed CNC cutters are fully utilized to evaluate the tool wear in the data-driven method. In this study, the sensitive features related to tool wear are explored from three aspects: time-domain, frequency-domain and joint time-frequency domain, and the feature extraction methods include time-domain statistical analysis, fast Fourier transform (FFT) between time-domain and frequency-domain, and wavelet transform (WT) in time-frequency domain. In this paper, the decision tree will be used for regression problems, rather than classification issues, to assess the tool wear value. And then, the AdaBoost algorithm is introduced to improve the performance of the decision tree regression (DTR), and the performance of the adaptive boosted decision tree regression (DTR-Ada) model and the original model are compared at the aspects of the accuracy, steadiness and applicability. The result shows the DTR-Ada model can improve the accuracy and stability of the fitting and prediction, and it also achieves a good effect on the applicability of the new tool wears prediction.

Abstract:As an important subject of the network, the behavior analysis of users is an important means to grasp the network security state and has a significant meaning on potential threat mining and early warning. Considering that users of similar roles in the power intranet have similar behaviors, this paper describes the behavior of individual users based on time sequence and builds behavior relevance among the users by self-learning of the similarity of users' behaviors to achieve abnormality analysis by means of behavior similarity deviation. Meanwhile, changes of users' basic attributes are considered to achieve abnormality early warning judgment. Simulation experiments show that the method can discover abnormal behavior and perform early warning by effectively using the similarity analysis of the behavioral sequences.

Abstract:The performance of the speaker recognition system degrades drastically in the noisy environment. A robust feature extraction method for speaker recognition is proposed in this paper. Warped filter banks(WFBS) are used to simulate the human auditory characteristics. The cubic root compression method, relative spectral filtering technique(RASTA) and the cepstral mean and variance normalization algorithm(CMVN) are introduced into the robust feature extraction. Subsequently, simulation experiment is conducted based on Gaussian mixes model(GMM). The experimental results indicate that the proposed feature has better robustness and recognition performance than the mel cepstral coefficients(MFCC) and cochlear filter cepstral coefficients(CFCC).

Abstract:With the development of space technology, more and more information needs to be computed on the satellite. The distributed technology of satellite cluster hence has become a research hotspot in recent years. Different from the ground environment, the satellite is limited with some constraints such as volume, power consumption and space radiation. The key technology such as architecture design, operating systems and resource management should be grasped to establish distributed environment on satellite cluster. To solve problems above, this paper first studies the distributed computing architecture and storage architecture on satellite cluster. Then, it focuses on the implementation of distributed resource monitoring on satellite cluster. Finally, it proves its feasibility and superiority through the authentication system of satellite cluster.

Abstract:The conventional laser pen and the storage device are both very small, so they are easy to get lost. To solve this problem, the paper proposes a teaching assistant based on cellphone. The teaching assistant communicates with the personal computer through WIFI, which makes fast-data-transmission come true. The cellphone can control the operations of the PPT, audio, video and test editing. The system sets up on the mobile terminals, and has no special demand on hardware. So, it is easy to use, and in line with the development trend.

Abstract:The three-dimensional reconstruction method based on multi-view image motion structure restoration is quite time-consuming and has little robustness. Aiming at these issues, a new three-dimensional reconstruction method for physical geological specimen based on Kinect is proposed in this paper. First, the GrabCut algorithm is used to extract the foreground objects, combining with the original depth images to generate the point cloud corresponding point of view. Then, it uses RANSAC algorithm to perform point cloud alignment crudely based on the SIFT feature. Next, the ICP algorithm is improved by introducing the methods of rejecting outliers and adjusting the weight dynamically to perform fine registration of point cloud. Finally, a complete three-dimensional point cloud model is constructed. The experimental results demonstrate that the proposed method can reconstruct the three-dimensional point cloud of physical geological specimen quickly, excellently and robustly. When the specimens lack structural features, the proposed method can handle it effectively.

Abstract:The dam's dispatching plays an important role in the utilization, protection and management of water resources. The dam is used as the basic assessment unit to build dispatching and management ability evaluation index system of a single dam, to build the model based on BP artificial neural networks for dam's dispatching and management ability evaluation. Finally, taking main dams on Huaihe river for example, the model is applied to verify samples, and the results show that the dam's dispatching and management ability evaluation model based on BP neural networks is a reasonable and feasible forecasting model.

Abstract:In view of the particularity of intelligent mobile applications and its data offline synchronization problems, we put forward a scheme of data synchronization, using JSON technology to design data exchange protocol and the SQLite database to store the mobile terminal offline data, using the conflict detection algorithm based on time stamp to improve the accuracy of synchronization as well as the incremental synchronization mode to ensure the efficiency and accuracy of synchronization. This proposed method is applied to an intelligent security system, and the results show that the efficiency of offline data synchronization based on JSON is about 8% higher than that of the traditional XML based scheme.

Abstract:Recently, the faster R-CNN has demonstrated impressive performance on various object detection benchmarks, and it has attracted extensive research interests. We train a faster R-CNN model on the WIDER face dataset with the ZF and VGG16 convolutional neural network respectively, and then we test the trained model on the FDDB face benchmark. Experimental results demonstrate that the method is robust to complex illumination, partial occlusions and facial pose variations. It achieves excellent performance in detecting unconstrained faces. The two kinds of network have their own advantages in detection accuracy and efficiency, so we can choose to use an appropriate network model according to the actual application requirements.

Abstract:The pipeline is the key technology of manufacturing high-performance CPU. The OR1200, which has been widely studied currently, is a 4-stage pipeline CPU with a free open source. Without MEM stage which should be designed in OR1200, the pipeline will be stalled to wait for load or store instruction. In this research, we design a MEM stage for OR1200 in LSU. Hazard detection and data forwarding units have been included for efficient implementation of the pipeline. On the other hand, when a data requested by a load instruction has not yet become available, it leads to load-use hazards. To resolve this hazard problem, we design a data valid signal Tag to control stalling of pipeline. The pipeline is stalled by the Tag signal for one stage and then continues with the forwarding of data, as the simulation result shows.