Aiming at the problem of unreasonable resource allocation due to incomplete information of hierarchical design in the downlink of multi-user MIMO-OFDM ( Multiple Input Multiple Output-Orthogonal Frequency Division Multiplexing) system，a cross-layer resource allocation algorithm is proposed to maximize the system throughput with the constraint of each user's data rate． The algorithm maximizes the system throughput with given transmission power and users' minimal rate constraints，it implements the joint subcarrier and power allocation considering the QoS( Quality of Service) constraints of different users，the queue state of MAC( Media Access Control) layer and the physical layer's channel state information． Simulation results show that the proposed algorithm can satisfy the QoS requirements of different users，achieve high system throughput．

To optimize ICI ( Inter-Carrier Interference) of spectrum pooling system based on wavelet，the ICI suppression optimization algorithm based on genetic algorithm is proposed． The genetic algorithm is used to get the optimum weight vector for suppressing ICI energy of un-LU ( un-Licensed User ) ． The algorithm can optimize un-LU interference energy without sacrificing the band-with of un-LU． Numerical and simulation results show the data transmission rate of un-LU is ensured，the ICI of LU ( Licensed User) is reduced and the BEＲ ( Bit Error Ｒate) performance of LU is improved by implementing the algorithm according to target BEＲ of un-LU．

In order to effectively detect the SBS ( Styrene-Butadiene-Styrene Copolymer) content in modified asphalt，an improved method based onInfrared spectroscopy is presented． By comparing the SBS modifier’ sspectroscopyandmatrix asphalt's spectroscopy，the characteristic absorbance peak at 966 of SBS modifier is used to detect the SBS content． Compound Simpson formula is introduced to calculate the peak area． Then，according to Lambert-Beer's law，the standard curve of the relationship between peak area ratio and SBS content are obtained by least square method． Finally，modified asphalt with different SBS contents are prepared from 70# matrix asphalt of Liao He Petrochemical Company． The average error of calibration samples and test samples is 0． 14% and 0． 02% respectively． The results show that the proposed method has the advantages of fast speed，high precision and easy operation compared with the original methods such as GPC． It provides a scientific and effective method for timely detection of asphalt SBS content in road construction projects in the future．

In order to solve the problem of low recognition rate of face recognition algorithm in face images with different pose，an improved algorithm based on convolutional neural network is proposed． The algorithm optimize the classical convolutional neural network LeNet-5's convolution kernel number，convolution kernel size，pool layer mode，regularization method，The pose-varied Face's feature can be extracted effectively． The experimental results show that the algorithm improves the rate and recognition time significantly．

The distributed optical fiber positioning system in coal mine usually adopts the dual Mach-Zehnder interferometer distributed optical fiber vibration sensing technology． The sensing distance is short and the spatial resolution of the system is not high． A distributed optical fiber vibration sensing technique for asymmetric double Mach-Zehnder interferometer is proposed． It uses two DFB ( Distributed Feedback) laser beams and DWDM( Dense Wavelength Division Multiplexing ) ， the noise reduction is processed by SVD ( Singular Value Decomposition) algorithm，and a cross-correlation delay algorithm based on SVD is proposed． Experiments show that the localization method can effectively improve positioning accuracy and sensing distance． When the sensing distance is 61 km，the positioning error is 51． 2 m． When the mine disaster happens，it can locate the trapped people accurately，and guarantee the effective rescue time．

In order to realize the rapid classification of geomorphologic type，based on the previous research results，using DEM ( Digital Elevation Modeldata) ，the process of extraction of landform information is made．According to the algorithm model，the optimal window of macro topographic factors is obtained by means of mean variation point analysis，eight topographic factors that can reflect the landform information are extracted． The dimensions of each terrain factor are unfied． The correlation coefficient matrix and the optimum combination of topographic factors are obtained by using snow entropy method． The classification of geomorphic types is realized through non-supervised classification by ENVI ( The Environment for Visualizing Images) software． Based on the secondary development platform of ENVI，adopting IDL ( Interactive Data Language) language to implement
program，complete automatic ( half) intelligent and quick division of landform types is realized． Taking the changbai mountain as the research area，from the macroscopic and microscopic classifying landscape，the result of evaluation and analysis of classification results are good． The realization of the system is of great practical significance to the processing and integrating geomorphic types and to the wide range of landform mapping in China．

In view of the poor practicality of the constant scheduling model for distributed power supply site selection and constant volume configuration，considering the time-series and uncertainty of DG ( Distributed Generation) output and load demand，based on the total investment cost of the distribution network system，total voltage deviation and system network loss minimization，multi-objective scheduling model is established． Firstly，the Monte-Carlo algorithm was used to simulate the wind-load scenario all year round，and the K-means clustering method is used to reduce the year-round scenarios． To solve the problem that the number of clusters in the K-means algorithm is difficult to set，the contour factor is introduced and improved to get the optimal number of clusters． Finally，the NSGA-I and non-biased strategy are used to optimize the process． In order to verify the
validity of the method，taking the IEEE33 node distribution system as an example，the comparison experiment with the standard genetic algorithm validates the validity and superiority of the proposed algorithm．

In order to improve the precision of seismic wave impedance inversion，a hybrid particle swarm optimization is proposed． This algorithm have the chaotic and genetic ideas． Chaos is introduced at the beginning of the search，making the algorithm have the ergodic property． Selection，crossover and mutation are introduced at the process of particle update and its convergent speed is faster and easy to jump out of local extremum． The algorithm is tested by model and is contrasted with ordinary particle swarm optimization． The result shows that the algorithm is better than the traditional particle swarm optimization． For the model with 15% noise，although has some errors，the coincidence degree is still good． The result shows that the algorithm has anti-noise ability． Finally，the algorithm is applied to the actual seismic data and it also has a good effect，the result shows that the algorithm has practicability．

In order to accurately report the information in the process of municipal facilities inspection，a municipal information management system is proposed based on Android platform． Through calling the camera and loading the GPS( Global Positioning System) ，Network position module in Android device，packing the data combined with the image，position information，other engineering information，are uploaded to the background server，and saved to the database． Through accessing Web pages on the browser，the background data management，task distribution are lmplemented． Test results show that the system can effectively extract the image and position information，upload the data and simplify the background workflow quickly and accurately，the system has a certain practical value．

To overcome the shortcomings of traditional approach to teaching quality evaluation of university，a more objective and reasonable intelligent evaluation method is proposed． According to the shortcomings of linear and static properties of traditional evaluation model，an implementation scheme of intelligent evaluator is proposed using ANN ( Artificial Neural Network) ． The training sample set of the intelligent evaluator is constructed with the data produced by traditional model． The 3 sub-evaluators are integrated with boosting to compose the intelligent evaluator，and the running rules of the evaluator is described． It is verified by practice that the proposed method better reflects the complex and dynamic relation between factors and results，and
produces good evaluation results．

Motivated by the recent success of visual representation and the problem of conventional approaches which were too correlated or too strict constraint，this paper proposed a framework of decorrelating mixed norm multi-task sparse learning representation to seek an effective and intermediate feature representation approach．By leveraging side information about multiple features to group all kinds of features，L2，1and L1 mixed-norm regularizations were used as penalty function，for L2，1-norm's feature sharing information among a feature group and L1-norm's sparsity and competition among unrelated feature group． This framework simultaneously enforced multi-features joint and considered the complementation among decorrelated features． Extensive comparison experiments on data set show that the proposed method can obtained idea sparsity and is competitive to the stateof-the-art methods for image classification，which demonstrated the efficiency of L2，1 and L1 mixed-norm algorithm
for extracting key information of an image．

To solve the problem of multilingual knowledge organization in cross language information retrieval，acording to the core idea of ontoterminology that the concept has stability independent of linguistic diversity，terms in different linguistic or cultural backgrounds can represent the same concept，combined with the needs of the education field，a standardized ontology model supporting multi language retrieval is constructed based on the thesaurus of education． Hoist the retrieval of terms to the conceptual level to eliminate the ambiguity of the interconversion of different languages，properly solving the problem of cross language semantic information retrieval in education field． Finally，we used Jena API to query the ontology and build a cross language semantic retrieval prototype system． The results show that the proposed idea is feasible．

For the large number of clustering objects with more dimensions and more complex，the system clustering is more complex． The fuzzy comprehensive evaluation clustering method clustering results are not accurate enough，the number of which is difficult to control． For this reason，a system clustering algorithm based on fuzzy comprehensive evaluation is proposed． This method performs fuzzy comprehensive evaluation on complex objects or objects which are restricted by many factors，extracts the whole features of the objects，and uses system clustering to cluster them． The validity of the algorithm is verified by clustering analysis of the test scores of five classes． Experimental results show that this method can effectively improve the accuracy of clustering，integrity，maneuverable，simplicity and so on．

With the rapid development of informatization education，personalized learning system has the advantage of breaking time and space limitation，attracting many learners of different ages． As an important part of the personalized learning system，domain knowledge model has attracted more and more attention from domestic and foreign scholars． Main modeling standards，and domain knowledge model was proposed to construct the reference standard． With the learning theory as the instruction，focusing on learners' individual behavior，students jointly participated in the collective wisdom of generation，in the important role of building a domain model，built by the individual domain knowledge model，the individual domain knowledge model aggregated into collective domain knowledge，realizing the connection between knowledge and knowledge，collective domain
knowledge model and affect the individual domain knowledge model，the optimization of individual domain knowledge model，and then complete the construction of domain model and optimization，the“personal”and “collective”the construction of knowledge system has a strong reference and guiding significance．

In order to solve the simulation problem of motor drag experiment，based on the idea of virtual simulation，using Unity3D technology，the motor drag simulation system based on Unity3D was realized． Through experiments，the system can accurately simulate the response of various electrical devices under different connection deployment conditions． The user can use this system to achieve initial experience of learning and simulation of the motor drag experiment． The system can complete some experiments that are not limited by time and space，which helps to improve the efficiency，accuracy，and safety of experiments，and reduce the cost of experiments．

The existing OpenCV template matching can realize the positioning of the screen printer，but the template image can not get the matching result in the search chart，and the positioning actuator uses the traditional direct value transfer method causing the positioning actuator stuck． In order to solve the above two problems，this paper proposes a template matching method based on similarity． It is verified that the method can solve the problem that the template image can not get the matching result in the search graph． The experimental results show that the similarity recognition method based on similarity is more than 91%，and the average time is less than 250 ms． A speed-matching movement method has also been proposed． This method solves the problem of the positioning actuator performing the stalling，the experimental results show that the speed-matching movement method does not exhibit a stalling phenomenon during 250 operations．