Search

In this paper, we propose a method for generating undulatory gaits for snake robots. Instead of starting from a pre-defined movement pattern such as a serpenoid curve, we use a Model Predictive Control (MPC) approach to automatically generate effective locomotion gaits via trajectory optimization. An important advantage of this approach is that the resulting gaits are automatically adapted to the environment that is being modeled as part of the snake dynamics. To illustrate this, we use a novel model for anisotropic dry friction, along with existing models for viscous friction and fluid dynamic effects such as drag and added mass. For each of these models, gaits generated without any change in the method or its parameters are as efficient as Pareto-optimal serpenoid gaits tuned individually for each environment. Furthermore, the proposed method can also produce more complex or irregular gaits, e.g. for obstacle avoidance or executing sharp turns.

With the development and maturity of computer and multimedia technology, computer-aided instruction represents advanced teaching ideas and methods. Based on this, the design of computer-aided course teaching control system based on supervised learning algorithm is proposed. The hardware composition and software design of the system design are introduced. Through the main modules: login and registration module, learning mode selection module, learning content selection module, test module operation. Relying on the constraints of multiple identities in the teaching process by the logical structure of the database, the teaching management platform can control and supervise student learning. In order to better test the effectiveness of the system designed in this paper, a comparative experiment of system control learning and free learning is performed. The results prove that compared with free learning, the learning of computer-assisted curriculum teaching control system can better urge students to learn, and the learning effect is better. The teaching control system is superior to the traditional teaching system in improving students ‘interest and motivating students’ learning enthusiasm.

With the continuous progress of Internet technology, more and more people are getting information through the network, so the number of cloud media is increasing, which brings great hidden danger to the network security. However, the traditional training management platform is difficult to deal with massive data, which leads to the imbalance of defense knowledge system in the platform and the instability of platform feedback. Therefore, a new network security defense knowledge training management platform is designed. By setting the balance constraint index, optimize the load balance layout of the training management platform; build the service framework of the training management platform, obtain the training user clustering results; set up training courses, realize the design of the network security defense knowledge training management platform. The test results show that compared with the traditional design platform, the stability of the designed platform is higher under the condition of massive network security defense knowledge. It can be seen that the designed platform is more in line with the basic requirements of network security defense.

In this paper, a method of palmprint segmentation and location is proposed. The proposed method focuses on region of interest (ROI) extraction of palmprint images which involve transition and rotation. Firstly, binary of palmprint image is used to define the edge of palmprint. Then we separate the fingers and palms and find the two valley points of the index finger and middle finger, ring finger and little finger. Finally, rotate image based on the two valley points and correct image position and create coordinate system according to valley points to determine ROI. This method provides a necessary preprocessing for further feature extraction and matching.The effectiveness of the proposed method is verified using the PolyU palmprint database.

As the origional Fast Hessian which is the most efficient blob feature detection algorithm can not meet requirements of those images in real-time applications to the target recognition,target tracking and so on,an accelerated Fast Hessian multi-scale blob feature detection algorithm is proposed to upgrade the detecting speed of the Fast Hessian.The basic idea of the proposed algorithm is to decrease the number of filter operations and to calculate selectively the values of sample points in the first and last scales for each Octave.Compared to the original one which calculates all the sample point values in these scales,the number of filter operations are distinctly decreased and the consuming time of detecting processing is also reduced.The experiments indicate that the accelerated Fast Hessian algorithm and the original one have the same detection results,but the implementation speed of the accelerated Fast Hessian is upgraded nearly 40% of the original one.It concludes that the accelerated algorithm is much more fit for real-time applications.

In order to improve network's security, a minimum cost of network hardening model (MCNHM) based on attack graphs is presented. Firstly, the bidirectional-based search strategy is used to search the network vulnerabilities’ relationship, which improves the generation efficiency of attack graphs, and reduces the system resource consumption; Secondly, this model gives the formal definition of minimum-cost of network hardening; Finally, it combines attack graphs and genetic algorithm, and transforms the problem of minimum cost of network hardening to a non-restraint optimization problem with penalty by establishing the corresponding mathematical model, which guarantees the network security with the least cost. This model is an important component of the National High-Tech Research and Development Plan of China, under Grant No.2009AA01Z432, a great of experimental results show that this model can find the minimum cost of target network, so it can help network security managers carry on safety protection in pertinence, and has important practical significance.

Because the existed approaches to harden networks have an unavoidable exponential worse-case complexity, and are not scalable to large networks, this paper proposes an optimal network hardening model (ONHM) based on parallel genetic algorithm by combining attack graphs and parallel genetic algorithm. Firstly, we describe the optimal network hardening problem; secondly, we establish a corresponding mathematical model, which converts the optimal network hardening problem to a non-restraint optimization problem with penalty. Through a large number of repeated laboratory tests, the experimental results show ONHM can find the optimal network hardening, and can be applied to large-scale networks.

From the perspective of circular economy, this paper discusses the choice of electronic-products reverse-logistics servicer. First of all, according to the characteristics of electronic products' recycling and reuse, an evaluation index system is designed to select its reverse-logistics servicer; Then, the index weights are determined with factor analysis method based on abundant investigation; Finally, the fuzzy comprehensive evaluation method is used to evaluate and select reverse-logistics servicer of electronic products.

A method to compress aeroengine test data is discussed in this paper. Three characteristics of test data are analyzed firstly. They are limited value range, limited variation range and limited value precision. Based on the characteristics, four processes including reducing decimal digits, eliding sign bit, eliding repeated value and eliding redundant coding bits are utilized to compress the data. Additional information needed for compression and decompression are discussed too in this paper. Two compression examples with real aeroengine test data show that the method can compress test data to very low compression ratio.

A new gray clustering analysis algorithm, which adopted cloud model as whiten function, was proposed to solve the subjective problem in power transformer condition assessment. The algorithm determined condition classification and parameters through their prior knowledge. The algorithm first optimized indicators of power transformer condition evaluation and stratified them. Then, cloud model was introduced as whiten function of gray clustering because the combination of these two methods could comprehensively consider randomness, ambiguity and gray of uncertainty problems. The parameters of cloud models determined according to prior knowledge of power transformer condition evaluation indicators, so it reflected the power transformer condition more really and objectively. Improved weight coefficient gray target theory analyzed dissolved gases and improved the accuracy of the model. Indicators polarities of the improved weight coefficient gray target theory were optimized so that it could evaluate normal state. For samples only with dissolved gases data, improved weight coefficient gray target algorithm can assess condition alone. Example indicates that condition evaluation results of the power transformer based on gray clustering with cloud whiten function consistent with the actual situation, and raise the scientific and objective of power transformer condition evaluation.

BP artificial neural network based on genetic algorithm training method was proposed in this paper to solving Renminbi (RMB) number recognition and artificial neural network BP algorithm problems in order to satisfy the technical requirements of financial sector sorter. The BP artificial neural network based on genetic algorithm training method combined the global search ability of genetic algorithm and local search ability of BP algorithm. Genetic operators such as selection, crossover and mutation were studied and chosen according to the object. Experiments show that RMB number recognition based on genetic algorithm and BP artificial neural network improves recognition accuracy and speed, and provide a new means of RMB number recognition.

Template matching is a method widely used in the area of target detection and tracking. In this paper, a new target match recognition method based on contour area feature is presented. After introduced the idea of this method, the method to construct the matching reference template and compute the matching comparability is explained. The matching performance of this method is analysed according to matching experiments with forward looking infrared real time images. The result shows that this method has applied value to forward looking infrared target recognition.

Infrared imaging is the concentrative embodiment of the highest level of contemporary infrared technology, and infrared imaging technology becomes one of the key technologies of current national defense science and technology. In the infrared imaging simulation, after gaining the temperature of object scene, firstly establish the radiation - temperature model to calculate the radiation quantum of object, and then produce the infrared images according to the gray mapping model. In addition, when calculate object real temperature with the measured temperature from thermograph, temperature value need to anti-count with radiation - temperature model. This paper gives an optimization algorithm by least-squares fitting in allusion to the calculating complexity by Planck formula. By comparing performance, it is proved that this algorithm can greatly reduce the calculating time in precondition of guaranteeing the accuracy.

Digital levels is typical product which using image processing to compete position measurement, but the encoded in digital levels is one-dimension position encoded. A surface positioning method based on two-dimension encoded recognition is presented. The structure of pen-type coding image detection system is provided. In order to complete position encoding the non-continuous code area coding and incremental coding methods are put forward and analyzed. In this system, we get position coordinates by the process of coding image. The image process steps and recognition method are discussed. This method can be applied to the development of interactive electronic whiteboard and computer input device.

Offline Chinese characters segmentation is one of the most difficult problems in Chinese character recognition, because handwritten Chinese characters have deformations, connected strokes and overlapped characters, and they often occurre with punctuations and digital numrbers. The multi-step offline handwritten Chinese characters segmentation method based on adaptive genetic algorithm was put forward in this paper to segment connected or overlaps characters, punctuations and digital numbers. Genetic algorithm chose the optimal threshold of projection profile histogram method to segment character string roughly. Genetic algorithm parameters were adaptively chosen according to different character string images. Then, an estimation strategy determines whether the character block to be merged If the character block contained character, then it would be merged with neighbor character block that had the shortest distance to it. Finally, Viterbi algorithm re-segmented the insufficient segmented characters. Original rule 4 was modified and a genetic algorithm rules were put forward to delete redundant paths. Experiments on HIW-MW database shows that the new algorithm has correct segmentation rate of 74.55% which is higher than other two segment methods. The new algorithm can segment Chinese characters, punctuation and digital numbers correctly, and it is an efficient segment method.

RS232 interface executes the role of Transportation and Communication, which has became the important interface between MCU of embedded system and peripheral equipment. Because RS232 mainly work in bottom of communication protocol, so it is important to protect the infrastructure of RS232. In test, pulse double electromagnetic is pulled into RS232 data transmission lines by coupling clamp, simulated differential-mode pulse voltage, basing on the test data, it will get the damaged mechanism of RS232 interface, meanwhile presenting the Logistic model of injected voltage pulse and probability of damage to the port interface, and getting the performance evaluation of RS232 interface, moreover estimating the voltage pulse range of upper and lower bounds at the port in the normal and damage state.

I2C bus and RS-232 serial port are the important data communication interfaces between the MCU of embedded system and the peripheral equipment. The embedded system performance will be affected under complex electromagnetic environment by the improper disposal to the interfaces, which results in the low speed and reliability. According to the speed matching problems in current embedded programming, the paper designed two kinds of buffers, the double buffers and cycle single buffer, which separately solved the problems of the speed improvement of I2C bus and the serial transmission reliability of RS-232. The result of the experiment shows that communications of I2C bus and RS-232 serial port under complex electromagnetic environment can be of speediness and stabilization, and the ideal goal has been obtained.

Computational complexity of K-L transform is the bottleneck of traditional eigenface algorithm. Test face image was divided into 9 sub blocks to reduce dimensions and computational complexity. Different weights were given to different parts of image according to their importance at recognition stage. So importance of human face key parts was enhanced. Within-class average face was adopted instead of mix average face, because it could keep with-class information and enlarge differences between classes. Adaptive learning step BP-based network was adopted as classifier. Experiments on ORL and Yale face database show that the recognition rate reaches 95.62% and 93.33%. The increases are 7.74% and 14.28% respectively compared with traditional algorithm. Analysis demonstrates the proposed algorithm in this paper has less computational complexity, higher recognition rate, and more robust than traditional algorithm.

In a certain extent, the gas component and content dissolved in the transformer oil indicate the insulation aging and fault degree of the transformer. The fault diagnosis system of transformer based on gas chromatography applies the analysis method of gas chromatography to transform each of the concentration into corresponding electric signal. After data processing, the signal is transmitted into working station to form chromatogram and made a fault diagnosing with grey relational theory. The system can detect the gas component, content, and speed of gas producing. It can find out the potential fault in the transformer as soon as possible and forecast the development trend of fault to ensure safe and economical running of the transformer.