Aiming at the issue that the SINR ( Signal to Interference plus Noise Ratio) of cognitive radar waveform always needs to meet certain threshold in the real Cognitive radar system, establish a new launch system model of cognitive radar based on two traditional waveform design criteria. This model takes the mutual information between the target and the echo waveform as the objective function with the constraints of SINR and energy. This paper solves the model by mixed penalty function and gets the optimal waveform of cognitive radar.The simulation results show that the waveform can assign the beam energy to the band which clutter beam energy response is weak and the target response is strong, meets the requirements of cognitive radar. So it solves the issue of the threshold of the SINR. It also maximizes the mutual information between the torget and the echo waveform. The proposed waveform design method has high practical significance.

In order to estimate the DOA( Direction Of Arrival) accurately and quickly, a spectrum estimation method based on maximum likelihood algorithm and SQP( Sequence Quadratic Program) was presented. First,the maximum likelihood function for DOA estimation is deduced, and it is used to convert the problem from parameter estimation to a nonlinear function optimization problem. Then, SQP is applied to optimize the process of finding the solutions of the maximum likelihood estimator in order to estimate the DOA. The proposed method reserves the asymptotic unbiased estimation of the maximum likelihood estimator and comparing with the genetic algorithm or particle swarm algorithm, this proposed algorithm has faster speed optimization, and has higher
convergence precision.

In view of the adverse effects brought by the noise in mine on the signal detection, we put forward a digital phase sensitive detection optimization algorithm which does not need external same frequency reference signal. And from the simulation experiment system of PED, effectiveness of digital phase sensitive detection algorithm has been verified on the receiving part. The algorithm samples measue signal many cycles, then the calculated value is got by multiply accumulating and averaging the number of sampling and reference value which is stored in array. After calculating the cophase and orthogonal component, we get the result of signal amplitude.Supposing that the standard of receiving signal in PED receiving system is judged by signal dependency over 0. 6
in the environment, the least SNR received by the system is -58 dB. The method can improve accuracy rating of receiving useful signal to some extent. Furthermore, the receiving instrument is made more miniaturizing and is convenient to the benefit of the algorithm.

To solve the problem of high complexity and difficult implementation in hardware of LBP( Layered Belief Propagation) algorithm. A modified decoding algorithm is proposed based on LBP algorithm, a function f(x) is introduced in the first step in order to reduce its complexity. Specific parameters correction factor and offset factor are introduced in the second step in order to improve its decoding performance. The simulation results demonstrate that the modified decoding algorithm reduces the complexity and improves the decoding performance compared to LBP algorithm which facilitates the hardware.

In order to give full play to the SAR(Synthetic Aperture Radar) on the deep geological target precise detection and rapid analytical advantage, to further improve the resolution of ground target recognition, based on discussing the basic principle and data analysis method, polarization measuring, interferometry and differential interferometry for SAR technology are analyzed. The favorable and unfavorable conditions of ultra wideband SAR applications are summarized. The scope of application of this technology and the technical specification, provides a theoretical basis and technical guidance for the SAR in the digital geological science and applied in geoinformation science.

A simulation research is carried out with the veDYNA software for the yawing motion of the vehicle during a tire blowout on the expressway. Firstly, the evaluation criteria of motion vehicles during a tire blowout are determined. Then, in order to replace the driver to control the direction of the vehicle, a fuzzy PID(Proportion, Integral, Differential) controller is adopted to plan the steering wheel angel of the vehicle with a blowout tire. This control scheme combines the advantages of fuzzy control method with the advantages of traditional PID control method. It can adjust the PID control parameters for different working conditions and complex road environments, to adapt to the changes in parameters of the vehicle during a tire blowout. The simulation shows that the trajectory control of vehicles with blowout tire based on the fuzzy PID can make the vehicle avoid yawing, the vehicle can be driven back to the original path after a small offset. It proves that this control scheme has a better adaptive ability.

In order to adjust the flight attitude of the quad rotor, an attitude data processing joint algorithm based on Kalman and DMP ( Digital Motion Processing) filter was adopted, and PID ( Proportion Integration Differentiation) algorithm was used to control its flight attitude. Based on the hardware platform, the dynamic model of the aircraft was established and the parameters of Kalman filter and PID controller were debugged. The flying experiment shows that after using the control algorithm, the control system can quickly adjust the flight attitude offset to control the quad-rotor effectively.

For the transition resistance broken problem in actual operation, we describe the composition and working principle of tap changer. Compared with previous research papers, AC detection method is proposed.AC test device through several tests, test waveforms compared with current law has more advantages. AC method to test the transition resistance breaking wave of tap changer has a unique analyzing result and facilitate the analysis, which greatly improves the accuracy of fault diagnosis. It will have a great value in practice.

In order to improve the manipulator system whose mechanical model of tracking precision is uncertain, a robust trajectory tracking control method, which is self-adaptive and based on an operational space, was researched. And taking into account the presence of unknown external disturbances and other non-modeled
dynamics, GL(General Linear) matrix and its multiplication operator were introduced and the ability of fuzzy systems was applied to approximate any nonlinear function which could render feedforward compensation to the manipulator dynamics. Modeling errors and bounded disturbances were inhibited through robust controls. This method does not need to solve the dynamic model when manipulator is in the operational space. According to Lyapunov stability theory, the stability of the system was proved. Finally, a relatively high effect of trajectory tracking control was ascertained by the simulation experiment on the designed controller.

Overhauser magnetometer is a high sensitivity magnetostatic measuring instrument based on Overhauser effect. It has become a hot research topic in recent years, because it has high performance and is widely used. The working principle and hardware structure of JOM-1 Overhauser magnetometer are given and
some key parts such as RF(Radio Frequency) excitation circuit, signal processing circuit, working sequence,noise level and experimental results are introduced. Field tests indicate that JOM-1 Overhauser magnetometer has achieved basic functions such as data collection, display, storage, viewing and so on. 0. 14 nT sensitivity is obtained and initial signal to noise ratio can reach 96:1, which is three times of the JPM-1 proton precession magnetometer fabricated by our lab.

It is important to select appropriate value ordering heuristics, as it influences the efficiency of constraint solving deeply. In this paper, two of the most respected intelligent value ordering heuristics are researched in depth. One is look-ahead value ordering heuristic, the other is survivors-first value ordering heuristic. The efficiency of the two kinds of value ordering heuristics are compared under the normal situation and the adaptive circumstance. The results display that survivors-first is better than look-ahead under normal environment, but decline in efficiency under adaptive situation. So we can choose different intelligent value ordering heuristics to achieve better property in different enviroments.

In the processing of chirp signal in the frame synchronization, FRFT(Fractional Fourier Transform)transform or relevant matching mode usually leads to the problem of the large amount of calculation. It is difficult to achieve real-time processing in hardware platform like stm32. The paper studies on obtaining precise
synchronizing datum of signal frame in the condition of complex underwater acoustic communication channel. The simultaneous technique achieves coarse synchronization by detecting three single-frequency signals and realizes frame synchronization scheme of fine synchronization in dechirp compression processing method while dealing with Chirp signal. It adopts simulation test method to measure time synchronization disturbance of the frame synchronization scheme in different channels. Compared with fine synchronization scheme in the use of relevant matching mode and FRFT transform, dechirp processing method possesses less massive calculation deviation, although the performance is not that good. We utilize icroe2 platform and some other hardware to achieve frame synchronization scheme of dechirp compression, and verify the feasibility of the scheme by testing noise attenuation and water tank environment.

A new 3D face recognition method is proposed aming to speed up traditional Iso-geodesic method based on Iso-geodesic curves and features extraction. This method is mainly composed by the following 3 steps:generating a serious of equidistant Iso-geodesic curves by geodesic distance, extracting feature points on every curves of first step and using PCA(Principal Component Analysis) method and voting to final face recognizes.This method has higher recognition speed than the traditional Iso=geodesic method when at the same recognition rate under Face WareHouse dataset. The average recognition time decreased from 2. 55 s to 0. 624 3 s.

In order to measure the performance of several wavelet-based image fusion methods, an image fusion assessment method is used. Two types of source image are used to perform the experiments, four-level wavelet decomposition based method is used to assess the images. Low frequency coefficient of wavelet decomposition matrix is used to fuse images by pixel average value method. Low frequency coefficient of wavelet decomposition matrix is used to fuse images by pixel value method. Then the methods are studied by some experiments. The results show that the information entropy of maximum pixel value and average grads is greatest. The information entropy are 7. 206 5 bit and 5. 983 7*10^-5 bit for underexposure images, 7. 130 8 and 4. 794 3*10^-5 bit for out of focus images. The proposed method gives the best performance in the experiments.

In order to solve the problem of selecting the clustering number of remote sensing images and positions of center points, Proposed a PCVOS ( Parallelized Center Vector Optimized Selection Algorithm for Remote Sensing Image Cluster) which introduces fuzzy evaluation of the objective function and put forward an evaluation mechanism of chromosomes to improve the diversity of category and space division of clustering chromosomes is proposed. The MPI multi-process parallel technology is simultaneously introduced to speed up the running speed of the algorithm. The experiment shows that compared with the traditional K-Means, ISODATA(Iterative Self Organizing Data Analysis Techniques Algorithm) and ACDE ( Automatic Clustering Differential Evolution) algorithm, PCVOS can obtain a better clustering effect, and make full use of parallel resources to speed up the running speed of the algorithm.

To improve the accuracy of driver fatigue detection, presents an improved eye state detection method.Firstly, we use reference white algorithm for the processing of illumination compensation on original frame image, then segment the skin region from background based on color feature. Secondly, we use gray-scale integral projection method to segment face and eyes, combining two parameters that change in the distance between eyebrow and eyelash the change in the number of black pixels in the eye region to judge the state of eyes. If the eyes are found closed for 5 consecutive frame images, the system will give a fatigue alert. The results show that the use of two parameters is better than the use of a single parameter.

Because of single function and difficult control of the current unmanned aerial vehicle, a multifunctional quadrotor aircraft, based on PID(Proportion Integration Differentiation) control algorithm, is designed to realise real-time aerial photography transmission and GPS(Global Positioning System) locating in the core of Cortex-M3, controlling the posture of the quadrotor aircraft by the angle of yaw, roll, pitch, acquiring the data including flight posture, throttle range and position parameter, building a dynamical model. Added with the wireless transmission, the data measured practically are transmitted to upper computer, compared with theory data. Then the controlling parameter is adjusted to shorten response time, steady the flight in accordance with algorithm and accomplish plotting and saving with the sensors data. The result shows that the quadrotor can fly steadily with multifunctional control and the distance of real-time aerial photography transmission can reach 110 meters.

Because the PC based license plate recognition system has a high cost, heavy complex operation, security is low and many other defects, we proposed license plate recognition technology embedded monitor video license plate recognition and embedded together by median filter, Sobel edge detection, global thresholding,texture features based on the approximate location, character normalization, different classifier character recognition technology to form a complete license plate recognition system. The system is small, stable, practical and accurate. We conducted 900 patients with different license plate recognition experiment under different circumstances, the accuracy rate is as high as 97. 3%, and the system can achieve very high precision.

For reducing the dependence of nonlinear observer on the precision model, a non conventional NN(Neural Network) observer for nonlinear system is proposed. The neuro-observer is a three-layer feedforward neural network, which is trained extensively with the error backpropagation learning algorithm including a correction term to guarantee good tracking and bounded NN weights. Designing the neural network observer is using artificial neural network to identify the nonlinear parts of the system and using a Luenberger observer to reconstruct the states of the system. The Lyapunov direct method is used in order to ensure the stability of the proposed non-conventional observer. The proposed observer is applied to 2 degrees of freedom horizontal manipulator to evaluate its performance. The simulation results show that the state observation of uncertain systems can be solved by the method and it is suitable for the low precision model of the nonlinear system.

In order to overcome the adverse effects on the performance of classification of SVM(Support Vector Machine) model parameters in the random selection, based on chaotic CABC-SVM ( Artificial Bee Colony Algorithm of Support Vector Machine) parameters optimization method is proposed. CABC algorithm uses the Logistic chaotic mapping initialization population and tournament selection strategy, further improves the artificial bee colony algorithm convergence speed and optimization precision, the classification accuracy as the fitness function when using the algorithm of penalty factor and kernel function parameters of SVM was optimized. UCI standard data sets experiments show that CABC algorithm has strong local and global search ability, the optimization of SVM can largely overcome local extremum points to obtain a higher classification accuracy, and can effectively shorten the search time. The method was applied to actual gear fault diagnosis, energy wavelet was used as the feature input SVM, classification accuracy rate reached 99. 4%, verified the feasibility and effectiveness of the improved method in this paper.

In order to enhance the accuracy and stability of lane detection, a lane detection algorithm based on bi-directional sliding window technology is proposed. Firstly, a combination of EDF(Edge Distribution Function) and Hough transform is used to obtain the linear models of lane boundaries. Secondly, bi-directional sliding window feature extraction technique is applied to detect real lane markings. Finally, a linear-hyperbolic lane model is used: in the near vision field, a linear model is used to obtain robust information about lane orientation; in the far vision field, a hyperbolic function is used, the curved parts of the road can be efficiently detected.Experimental results indicate that the proposed method can enhance the adaptability to deal with the random and
dynamic environment of road scenes, such as curved lane markings, sparse shadows, object occlusions and bad conditions of road painting.