To resolve the problems of high complexity of sub-pixel interpolation operation and large access volume of storage in H.264/AVC standard, a kind of sub-pixel interpolation operation is presented. It replaces the 6 order filter with a 4 order filter which is easy for hardware implementation of sub-pixel interpolation. Based on the algorithm, a kind of 1/4 pixel precision interpolative pipeline architecture for 8×8 basic block is proposed. It is indicated by the performance analysis and filter structure that the structure can complete 32 interpolation operations of 1/2 pixel location in one clock period, which can be applied to data block with a variety of size. It has the characteristics of small area and fast speed. The experimental results show that compared with H.264 standard, the new algorithm is able to reduce space complexity by 15%, improve PSNR(Peak Signal to Noise Ratio), reduce bit rate and improve the performance of coding.

The bandwidth limitations and the low dispersion value of conventional dispersion compensating optical fiber make it not conducive to production of the dispersion compensating optical fiber. In order to solve this problem, using the plane wave method, the dispersion curves of fibers with different structures have been obtained, which gives the theoretical guidance for the production of dispersion compensation fiber. Results show that for the ordinary photonic crystal fiber with hexagonal structure, the dispersion value for short-wave-length optical signal can be reduced and the long one can be increased by reducing the air hole diameter or increasing the air hole interval. Large negative dispersion value can be obtained at long wavelength by using the dual-core structure, and flatten the dispersion curve if reduce the distribution area of the air holes, the fibers with flat dispersion curves can be used as a broadband dispersion compensating fiber.

A space vector modulation method is designed for eliminating harmonic pollution and improving power factor. Based on fixed switching frequency, this SVPWM (Space Vector Pulse Width Modulationis) is able to keep an uncontrolled maximum current. The control principle of this SVPWM is disclosed, and mo
dules for simulating algorithm and control are designed with Matlab. Results of simulation tests have shown that this modulation method can improve the
 utilization rate of system's DC (Direct Current) voltage and reduce the switching loss, demonstrating a good dynamic quality and a reduced total harmonic distortion.

In existing energysaving methods for heterogeneous wireless networks, the potential for inter-network collaboration has not been brought into full play, lack of the comprehensive application and theoretical analysis of carrier frequency turn-off technology and base station turn off technology result in unsatisfactory energy-saving effect. In order to solve the problem, this paper proposes a cooperative ESDT(Energy Saving Algorithm Based on Dynamic Thresholds)  in heterogeneous wireless networks. Aiming at minimizing the overall system energy consumption, the algorithm combines carrier frequency turn-off tech nology and base station turn-off technology, and applies queuing theory to adjust thresholds and dynamically turn on/off multi-granularity resources acco rding to system load condition to achieve energy-saving effect. Simulation results show that the proposed algorithm compared to the algorithm dynamically switching the carrier frequency, has better energy-saving effect and higher resource utilization on the premise of the guarantee system performance.

In order to resolve the problem of cognitive radio spectrum sensing with continuous-time measurement, a spectrum sensing method based on the pu
rely discontinuous Markov process was proposed. The channel state can be divided into idle and busy state. The sojourn time of the purely discontinuous Markov process follows exponential distribution and the famous Fokker-Planck equation, the channel state matrix, the primary user's accumulation sojourn time and the distribution condition of the primary user in the time domain can be derived. From the results of the simulation, this method can preferably perform the task of predicting of the channel state and the task of tracking the state of the primary users.

As a means of antiinterference, the channel coding has important research significance in many technical problems which can be resolved in the visible light communication. Cascade code makes short code to long code, which can satisfy the channel of yards long in the error correction process to get the same ability of long code error correction, and easy to implement with low complexity. We use convolution code and RS code to generate cascaded codes of the visible light communication channel coding. Through simulation test, this method compared with traditional convolution code, when the bit error rate reaches more than 10^-4, it can improve the coding gain amounting to more than 1.5 dB. The ability of error correction has more promotion. So this cascade code is more suitable for visible light communication requirements in practical application.

A digital high precision constant current driver power is designed for array QC(Quantum Cascade) lasers in the Midinfrared gas detection system. Time division multiplexing is applied in this system to drive several QC lasers which are series and parallel. TMS320F28335 is used as a control center and the constant current driver power has a frequency range of 1~10 kHz, duty cycle from 0.5% to 40%, current amplitude range of 0~4 A by using big power semiconductor devices. In the control scheme, analog PI(Proportion Integral) control algorithm is adopted to improve the precision and stability of the driving current by using double feedback mode of two level analog PI links. Considering of the security of QC lasers, the system is designed with a slow start circuit, over-current and over-voltage protection circuit, ESD(Electro Static Disoharge) protection circuit, etc. The results show that when working in a certain condition that the pulse amplitude is 2.0 A, pulse frequency is 5 kHz, the pulse width is 2 μs the stability of output current pulse is better than 1.7×10^-3 A, the stability of pulse width is better than 2.4×10^-2 μs, the pulse rise is less than 11.9 ns. and the pulse fall time is less than 9.6 ns. At last, driving several QC lasers for testing at the same time, the results show that there is no glitch regarding the driving current and the center wavelength of the tested QC lasers does not shift. Therefore, the fabricated driver provides superior perform
ance that guarantees the application of the QC lasers in infrared gas detection.

An improved static slicing algorithm based on reverse flow has been put porward, to solve the disadvantage of traditional slicing algorithm that spends too much time and space. The improved algorithm only by analyzing the control dependence, to obtain the overall framework of the program, then scanning from inside to outside in the way of expanding based on reverse flow of control starting from the slicing point, continually gaining data dependence associated only with the slices, followed by analysis of the program slicing. This makes the algorithm not only solve the shortcomings of the traditional slicing algorithms, but also reduce the workload of the calculation control flow, eliminating the need to calculate the predicate dependency set, to further improve the efficiency of the slices, to reduce the w
aste of resources.

In order to solve the problems in the characteristics of UAV(Unmanned Aerial Vehicle) image with large frame, i.e., large rotation angle, large difference in scales an
d color difference, a matching method named multi-scale LSM(Least Squares Matching) algorithm based on SIFT (Scale Invariant Feature Transform) features with epipolar and homography constraints, which can improve the matching success rate is designed. On the top pyramid images, SIFT image matching is done to obtain matching points. The homography matrix and basic matrix are calculated with the matching points by the improved RANSAC(RANdom SAmple Consensus) algorithm. And the harris feature extraction is used to obtain many feature points. According to epipolar and homography constraints two-dimensional concordance correlation coefficient algorithm is used to dense stereo matching. The homography matrix is updated for deleting false matching points by setting threshold. Corresponding image points are used to obtain sub-pixel accuracy by LSM. Based on three groups of comparative tests with actual aerial photograph images, i.e., images with large rotation angle, lager different scales and color difference, it is proved that this method is effective.

In order to improve feature point location affected by noise in stereo matching, an algorithm of facial feature matching based on phase congruency detection is presented. This method extracts the face region in the obtained parallel binocular images to reduce the search area of the stereo matching, and uses phase congruency for detecting the edge and texture feature in the face, test results are not affected by brightness, contrast and other factors, so the images with different lighting conditions can be used the fixed threshold to avoid the difficulty of threshold selection in feature detection. RHG(Rotation Histogram of Gradients） algorithm is utilized to extract the feature point descriptor for feature matching, minimizing the influence of change in viewpoint. The experimental results show that the algorithm has better stability and higher matching rate, the rate is above 80%.

In order to have a better understanding of latest trends and technologies of network flow collection, and have a better research on network flow collection and other related researches, we summed up the methods and representative tools of network flow collection. By summarizing and analyzing the methods and tools, illuminate the key problems and core technologies of network flow collection so that provide the basis for further research of network data used by network measurement and management and reference of improving efficiency, veracity, practice ability, and applicability, and look forward to creating more efficient methods.

To improve the efficiency and accuracy in choosing concepts and relations of domain ontology, we present an unstructured data based ontology learning model. In the process of extracting the candidate concepts for synthetic word processing, we modified calculation method of frequent item of association rules, and combined with a bitmap to store physically adjacent relationship between the terms after word processing. We filter candidate concepts by calculating areas correlation and areas consistent degree. The association rule credibility and hierarchical clustering methods were used to extract nontaxonomic relations between concepts and classification relationships. Experimental results show that this model is rational in the aspect of domain ontology learning and this algorithm is efficient and accurate in the aspect of extracting concepts and relationships.

To solve the problem of pixel expansion, a (2,3) visual cryptographic scheme without pixel expansion is proposed, in which the generated shares have the same size as the original image. Instead of considering one pixel of input image at a time, the proposed schemes consider 4 pixels at a time and generate 4 output pixels in each share. So the decrypted image remains the same size as the secret image. This technique generates shares with less space overhead compared to existing techniques and may provide better security. It is also easy to implement like other techniques of visual cryptography. The experiment results proved the efficiency of the proposed scheme.

In order to solve the problem of imprecision in Chinese webpage theme feature extraction, feature extraction algorithm for Chinese webpage theme is studied. Webpage theme feature extraction is the foundation of topic web crawler to calculate webpage correlation. Considering two classifications of theme webpage, we improved the commonly used text feature item weighting method of TF-IDF(Term Frequency\|Inverse Document Frequency). We combine Semi-structured characteristics of webpage, feature's position information, present a new calculation method of linear feature item weighting. This method can effectively improve the theme webpage recall rate and precision rate.

In the view of the code clone and software piracy, we analyzes three kinds of traditional similarity detection method based on program structure, then put forward a similarity detection method based on syntax tree of program. In the method, the source code is parsed and a syntax tree is produced for it. Then the similarity of source codes is calculated by analyzing the structures of their syntax trees. The senior copy means are eliminated from the angle of syntax structure. By using a set of plagiarized scripts as testing programs, the experimental results show that the similarity detection method based on syntax tree is effective to detect 10 kinds of plagiarism means code plagiarism, such as adding redundant statements and replacing control structures with equivalent structures.

In order to form a more convenient and stable wireless communication network (Ad Hoc network), which is convenient for communication between wireless terminal users based on the bluetooth, we discussed Bluetooth communication technology, among which OLSR (Optimized Link State Routing Protocol) is an active quick routing protocol. Specifically, the operating principle of OLSR was elaborated and corresponding algorithms were designed to realize the formation of Bluetooth wireless network. Real machine test result shows,  under the stable environment of the nodes, the network can be built, and implement the bluetooth to jump more communication between nodes.

For the stable communications of Ad Hoc network, we optimized the AODV-FABL(Ad-Hoc  On-demand Distance Vector Fast Adaptive Backup Local） protocol, which is an improved routing protocol(AODV-OFABL：AODV-Optimize Fast Adaptive Backup Local). The HELLO message of the AODV-FABL is modified. The delivery ratio of packet and nodes mortality were reduced. Based on the current army's structure and reality warfare, we establish the special warfare, field warfare and urban warfare model. Using NS3 simulation software, we obtained the best combation between the model with the route. The results show that AODV-OFABL protocol is more suitable for using field warfare. When stayed for 60 s, the delivery rate of packet reached 97%, with the average delay down to 320 ms.