In order to solve the resource distribution problem in the joint resource allocation model，a novel power control and channel assignment algorithm is proposed，which is named CCJPCA ( Congestion Control oriented Joint Power control and Channel assignment Algorithm) ． In the CCJPCA algorithm，the hybrid coding
strategy is used to achieve the co-evolution of the power and channel variable． The reward mechanism of Q-Learning algorithm is used to realize the adaptive selection of mutation strategy，which ensures the reasonable allocation of network resource． The simulation results based on NS-3 ( Network Simulator-3 ) indicate that CCJPCA algorithm can give priority to network resources of network bottleneck links， achieve algorithm convergence speed，reduce network queuing and retransmission delay，and lower the average packet loss ratio．

The important principle of Cognitive Ｒadio system is to guarantee the communication quality of licensed users． In order to improve the communication quality of licensed users，the inter-carrier interference between licensed users and un-licensed users must be effectively suppressed． A multi-objective optimization
based interference suppression algorithm for Cognitive Ｒadio is proposed． By analyzing the interference energy of the licensed users in the spectrum pooling，the algorithm locates the subcarrier sequence which produces the maximum interference to the licensed users in the un-licensed user's subcarrier． Under the premise of meeting the data transmission rate target of the un-licensed users，masking these maximum interference subcarriers effectively
suppress the interference energy of the system． The simulation results show that the new interference suppression parallel algorithm can obviously reduce the interference and make licensed users obtain better bit error rate performance．

In order to improve the extraction to the sensing of brillouin scattering spectrum of the accuracy of brillouin frequency deviation，the squeezing factor of using particle swarm optimization algorithm is used to adjust the weights of ＲBFN ( Ｒadial Basis Function Net) network． The proposed algorithm overcomes the shortcoming of traditional ＲBFN neural network which is easy to fall into local extremum． Using PSO ( Particle Swarm Optimization) algorithm after adjust the weights to the transmission network，precision of brillouin scattering spectrum are extracted，ensurimg the solution speed and precision． In the process of numerical analysis，a new algorithm is used to estimate the scattering spectra of different line width and different SNＲ( Signal Noise Ｒatio) at different temperatures． Through the experiment the brillouin scattering spectrum data are obtained． Using YSPSOＲBFN( Particle Swarm Optimization with Shinkage Factor Shirnhage Factor-Ｒadical Basis Function) algorithm to deal with the experimental data，the results show that the algorithm can improve the accuracy of brillouin scattering spectrum feature extraction，the fitting error is 1． 99 MHz，under 25 ℃ when temperature fitting error is reduced．In 85 ℃ the frequency shift of fitting error is 0． 047 MHz． Therefore，when the algorithm is applied to the scattering temperature and strain sensing system of brillouin，it has great application prospect in improving detection accuracy．

In order to measure the flight time of photons，and improve the distance of laser radar ranging and ensure the stability of minimum time resolution，A pixel-level high dynamic range 16-bit time-to-digital converter is designed． The resolution is less than 330 ps and the detectable distance is more than kilometers． This design uses Xfab 0． 18 μm CMOS process for tapeout verification．

In order to solve the problem of processing 1． 2 V digital signal in LVDS( Low Voltage Differential Signaling) driver with 2． 5 V power supply，due to the poor performance of traditional level conversion circuit and the problem of bit error，an LVDS interface circuit for CMOS ( Complementary Metal Oxide Semiconductor)image sensor chip is designed． The digital circuit is powered by 1． 2 V and the LVDS driver is powered by 2． 5 V． This paper proposes two level shifting circuit schemes to solve this problem． In the first solution，hysteresis comparator is used as the level shifter． In the second solution，the 1． 2 V digital signal is level shifted first，then the D trigger is used to sampling the shifted signal and the error code will be avoided． TowerJazz 65 nm CMOS technological process is employed for tape-out verification． After the tests，the problem of LVDS driver error code can be efficiently solved by these two solutions．

In view of problems encountered in the application of intelligent MOSFET ( Metal-Oxide-Semiconductor Field-Effect Transistor) in automotive electronic system，the corresponding technical solutions are put forward． Aiming at the damage problems of Smart MOSFET under the condition of power reverse，short circuit，over load and load dump，the hardware protection circuit and software control strategy are put forward．Aiming at the low current feedback accuracy of Smart MOSFET，which can not meet the control accuracy，the method of improving the sensing accuracy through calibration is put forward． The on-resistance is reduced by parallel Smart MOSFET to improve the current capacity． To increase the capacity of the clamping energy，a TVS (Transient Voltage Suppressor) is connected with smart MOSFET in parallel． The feasibility and validity of the schemes are verified by experiments and real automotive tests．

In order to improve the precision of the incremental photoelectric encoder，reduce the complexity of the signal processing circuit and improve the stability of the servo control system，a high precision incremental encoder with a single loop output of 32 400 periodic square wave signals is developed． The encoder has an outer diameter of 90 mm，and the encoder has 3 240 lines． The data processing system completes 40 times accurate segmentation of Moiré fringe signals． Experiment result shows that the orthogonality deviation is less than 15% and the uniformity deviation is less than 20%，which greatly reduces the jitter of the servo system when the incremental encoder is applied．

In order to meet the needs of the experimental teaching of television theory，the digital television principle experiment teaching platform based on Matlab software is developed． According to the teaching content of digital television theory． Six experimental modules are designed，each module contains different experimental projects，covering the important aspects of digital television signal generation，processing and compression coding prediction． The design principles of the experimental module are given，and the experimental principle and content of the module are introduced． The application of the specific experimental module is taken as an example to explain the use of the platform． The practical application shows that the digital television experiment teaching
platform meets the demand of experimental teaching and has achieved good teaching effect．

In order to improve the image quality of images and degrade the effect of uneven illumination ambient noise in the identity recognition based on finger knuckle print，it is necessary to enhance the image in the process of image preprocessing． A new method of finger knuckle print image enhancement is proposed based on guided image filtering，due to abundant texture features of finger knuckle print image． And effectiveness is validated through the experiment in the database of finger knuckle print． The results show that the effect of uneven illumination can be avoided and details of texture can be highlighted by using the method of image enhancement．Conclusions were given by quantitative study，and more detail information are preserved after the imageenhancement，which contribute to the accuracy of subsequent feature extraction and identification．

Aiming at the problem of moving object detection based on mixture of Gauss background model，mainly the false detection of light mutation and the“ghosting”of sudden moving object，an algorithm of moving object detection of mixed Gauss background model based on three frame differencing is designed． According to the proportion of the foreground of the image to judge whether the light is changed，the three frame difference method is used to divide the background area，the moving area and the exposed background area． According to the illumination，the learning rate of each region can be changed in time to adjust the update speed of the background of the mixed Gauss model，and a new method based on three frame difference and adaptive learning rate is proposed to update the background of mixed Gauss model． This method makes the new background model produced by abrupt illumination and sudden motion of the target rapidly updated，improving the detection result of moving objects in these two cases． The experimental results show that the moving object detection algorithm can avoid large area error phenomenon of illumination mutation，and solve the “ghost”problem of moving object．

In order to solve the problem of heavy workload and human error in industrial field pointer instrument identification，a method of industrial linear pointer instrument identification combined perspective transformation with Hough transform is proposed． After the image denoising and perspective transformation，Hough transform is used to detect the image features of the instrument pointer，and the actual pointer deflection angle is simulated with this image feature． Finally，the final number is calculated by this angle value，linear pointer instrument identification work is completed． When the frontal image cannot be well acquired in the field，recognizing industrial linear pointer meter readings is achieved well on some ways． In the new situation，a new way of thinking for the application of artificial intelligence in the industrial field is provided by this method．

In order to improve the searching performance of the algorithm in finding high-dimensional functions，an improved particle swarm optimization algorithm is proposed． This algorithm combines the good global search performance of AFSA ( Artificial Fish Swarm Algorithm) with the advantage of strong local search performance of PSO ( Particle Swarm Optimization) ． In initial period，AFSA was used to obtain the optimal population，and PSO was used to achieve the fine search in the later stage． In order to solve the problem of arbitrary initial population and uneven distribution，uniform initialization was used to optimize the distribution． Aimed at the poor global search direction and low efficiency of the algorithm，grouping strategy based on SFLA ( Shuffled Frog Leaping Algorithm) was adopted，and different search strategies for good individuals and other ordinary individuals in the group were used to improve the purpose and efficiency of search． Because PSO is prone to stagnation and results in low accuracy of the final result，the improved elite Gaussian learning was introduced to enhance the accuracy of the final result． The proposed algorithm is used to solve function optimization on six standard functions and compared with other algorithms，the results show the improvement is effective and superior to other algorithms．

Aiming at the problem of low efficiency and slow learning in discrete state of Q-Learning algorithm．The improved algorithm adds a learning process on the basis of the original algorithm，and makes deep learning of the environment．An improved Q-Learning algorithm is proposed to simulate in grid environment． It has been successfully applied to the path planning of a mobile robot in a multi barrier environment，and the results prove the feasibility of the algorithm． The improved Q-Learning algorithm can converge faster，reduce the number of learning，and increase the efficiency by 20%． The framework of the algorithm has strong generality for solving the same kind of problems．

The aim is to make people has a overall，clear and accurate definition about data quality dimension for facilitating an agreement at the field level，and also provides the basis for choosing appropriate framework from existing data quality control technical architecture． This thesis carries on the overall research of development of data quality，data quality dimensions and data quality control framework． By studying complex data quality dimensions，define their names and definitions to provide an overall data quality specifications and understand data quality dimensions correctly． By researching and contrasting data quality control technology framework，provides a basis for the reasonable choice according to the field requirements． The research results can guide the field to choose the appropriate data quality framework according to their own needs scientifically and realize the
assessment by overall data quality dimensions，it is helpful to reduce unnecessary manpower and material resources．

The grid connection of electric vehicles and distributed power supplies brings significant uncertainty to the grid． In order to make grid analysis closer to the actual grid． It simulates output through the probability density function of load and distributed power output． It proposes to use Latin Hypercube Monte Carlo simulation and radial basis neural network to deal with distributed power supply and electric vehicle probability model． This method takes full account of the randomness，intermittency and correlation of EVs and distributed power supplies． Using Latin hypercube MCS ( Monte Carlo Simulations) to compare with traditional Monte Carlo simulation methods，the sampling scale is reduced and sampling coverage is improved． The radial basis neural network is used to solve the power flow calculation equation，which avoids the calculation of Jacobian matrix and partial guide in the traditional method． Through simulation，the calculation results of the proposed algorithm in the improved IEEE14 and IEEE118 node systems show that while ensuring accuracy，the algorithm speeds up the algorithm greatly，and is suitable for the solution of probabilistic power flow in large-scale power systems． The improved IEEE118 node in the system，the running time is reduced by 99． 9% compared to the traditional MCS．

In view of the lack of pertinence in teaching plan and timeliness in teaching effect valuation，an early warning model of interactive teaching system is proposed based on rule detection and outlier detection． Analytical algorithms based on the learning process and academic performances are designed for automatic intervention and active intervention based on learning behavior analysis． After information visualization，the result shows that all parties involved can acquire early warning information intuitively and efficiently，which can promote the improvement of learning effect and learning success rate．

For public safety video image information trajectory tracking problem，a fast video content retrieval system based on image recognition is designed． The HOG( Oriented Histogram Feature Extraction) algorithm is used for face location，and the ＲesNet model is re-trained through migration learning to establish a dedicated Neural network classifier． The process of retrieving video content is to extract the key frames of the video，locate the characters therein，extract the feature values，and use the trained classification model for classifying． The classification information is marked on the picture，the relevant information is stored in the database，and the database is queried for detailed information． Experimental results show that the system can effectively locate the information of the video and accurately capture the image． In the field of public security，it has application prospects．

The algorithm of collaborative filtering is only considered to analyze the user-item evaluation matrix traditionally． The properties of the item or the user are often ignored． In order to solve this problem and improve the accuracy of the algorithm of collaborative filtering recommendation，we apply the attribute categories of the items into the formula for calculating the similarity of items． The specific method is as follows: firstly，get the degree of difference between the item properties by creating the items' attribute categories table．Secondly，apply the degree of difference between the item properties into the pearson correlation formula and calculate the similarity between items． The experiment results show that the recommended MAE of the improved method is smaller and the hit rate is higher compared to the traditional collaborative filtering algorithm．

In order to solve the problem of traffic police dealing with emergencies and day-to-day inspections，a rapid identification of vehicle license plates based on Android platform and 4G network transmission is proposed．The method uses a 4G network to transmit pictures taken by a mobile terminal to a remote server to identify the license plate number． Use the license plate number to query the violation information database，and return the query result to the mobile client in real time． As a mobile terminal，a smartphone uses JAVA to layout a UI( User Interface) interface and implement functions such as photographing，display，and information input． With a simple interface，easy operation，and complete system functions，the APP can effectively improve the efficiency of the traffic police in dealing with emergencies and daily inspections，alleviating the problem of traffic congestion caused by law enforcement．

A non-contact height measuring instrument using infra-red sensing technology is designed． This system takes SCM ( Single Chip Microcomputer) as controller，designs the infra-red transmitting and receiving circuit，at the same time uses the motor and its drive to cooperate with the infra-red for measuring height and completes corresponding software program． The experimental results show that the instrument has simple facilities and easy to realize． It also has the high accuracy and reliability and can be intelligent applied to non-contact height measurement．