For desert seismic data with low signal-to-noise ratio，random noise is mainly non-stationary and non-Gaussian low-frequency colored noise． It is difficult to effectively identify and characterize seismic signals by using conventional noise suppression methods，resulting in the loss of signals with weak textural structure． The fractional order gradient operator is introduced，and a spatial fractional complex diffusion based seismic random noise suppression algorithm is proposed． The spatial fractional order gradient is used to enhance the weak texture and details of seismic signals，and the diffusion intensity is controlled according to the structures of seismic events under the complex diffusion framework． The synthetic seismic data and field desert seismic data are used to evaluate the performance of the proposed algorithm． The experimental results show that the spatial fractional complex diffusion filtering algorithm can suppress random noise and retain the effective signal structure in desert seismic data．

In order to solve the problem of nonlinearity in high frequency data，the LMD ( Local Mean Decomposition) method is proposed to estimate the volatility of high frequency data． Firstly，high-frequency simulation data are used to verify the feasibility of the estimation method． Secondly，taking the closing price of Shanghai and Shenzhen 300 index at different frequencies as the research object，the volatility is estimated by LMD method，and the relative error statistics are calculated． The results show that the LMD method can effectively estimate the volatility of high-frequency data and solve the nonlinear problems in high frequency data．With the increase of sample frequency，the estimation accuracy gradually improves． The method of LMD provides a new idea for non-parametric estimation of volatility in high frequency data．

To improve the quality of reconstructed images，it is necessary to determine the decomposition scale K value and the effective intrinsic mode function in 2D-VMD ( Two Dimensional Variational Mode Decomposition) algorithm，a new algorithm combined 2D-VMD and CC ( Correlation Coefficient) is proposed and used to reconstruct image． Firstly，the image signal is decomposed into sub-modes of different central frequencies by 2D-VMD． Then，the CC value of the decomposed Intrinsic Mode Function ( IMF) and the original image function is calculated to determine the effective Intrinsic Mode component according to the CC criterion．Finally，image denoising is achieved by reconstruct with effective intrinsic modal component． Simulation results show that the decomposition scale K and the effective sub-modes can be obtained through combined 2D-VMD and CC，which has high image reconstruction accuracy and improved image quality．

An improved timing synchronization algorithm based on ZC( Zadoff-Chu) sequence is proposed in order to improve timing synchronization accuracy of OFDM ( Orthogonal Frequency Division Multiplexing ) system． This algorithm combined the sequence inversion of Minn algorithm with the sequence conjugate of Park algorithm． The new sequences replaced PN sequences with ZC sequences that good autocorrelation and low crosscorrelation．Even in the timing position error，the system can still be precisely timing synchronized． Under the channel of AWGN ( Additive Gaussian White Noise) ，the number of subcarriers is 256，16QAM modulation and demodulation，Matlab simulation shows that the improved algorithm eliminates the influence of minor peak on timing synchronization in Minn algorithm and Park algorithm． Timing deviation and MSE ( Mean Square Error) are lower than Minn algorithm and Park algorithm within － 15 ～ 15 dB，which improves timing accuracy． When the SNR is 15 dB，the timing MSE of the improved algorithm is 4． 2 × 10 － 6，which is the optimal state for estimating performance．

In order to solve the problems of poor information fusion effect and prolong information fusion in current methods，a multi-channel information fusion method for wireless networks based on grid system is proposed． The support degree of wireless network sensor is calculated by using the correlation function in the fuzzy theory． The knowledge element model is used to obtain the fusion set of information，and it is taken as an information unit to provide the basis for the multi-channel information fusion of wireless network． Based on the entropy weight method，the European distance measurement between the multi-channel information of wireless network is introduced． The global and local fusion weights of information are obtained by using the concept of distance entropy，and the global and local fusion results of channel information are calculated by using linear weighting method． Finally， the fusion of multi-channel information in wireless network is realized． The experimental results show that the proposed method is effective and feasible for multi-channel information fusion in wireless networks with short time delay and good real-time performance．

The spoken dialogue system has great practical value and broad commercial prospects，and has significant research significance for its implementation which involves many directions in the field of natural language processing． Dialogue management is the core technology of spoken dialogue system，and plays an important role． The evolution of dialogue management forms the main historical line of the development of spoken dialogue system． In recent years，the research in the field of spoken dialogue system，especially the research on dialogue management，has yielded fruitful results． In order to generalize and summarize the main work in this field，this article first introduces the various components of the spoken dialogue system，including the main research contents and key technologies． Then focuses on the core module of the system． The development and evolution of the technology are outlined． Finally，we have studied the further developments of dialogue management methods，and have forecasted the future develepment．

An accurate count and size recognition system of steel balls is developed in order to solve the problems
of inaccuracy of steel balls weighting and counting，steel balls spherical bruises，mixed balls during bearing steel
balls production process． A CCD ( Charge Coupled Device) camera is used in this system to get the images of the
steel balls in the counting-plate，to calculate the quantity of the steel balls and recognize the steel balls of abnormal
size． The system is composed of a feeding channel，an estimated count plate，an accurate count plate，a CCD
camera，a single ball compensating device and a computer etc． The steel balls scroll to the estimated count plate
through the feeding channel and they are arranged in one layer． The area of the estimated count plate can be
adjusted according to the dimension and quantity of the steel balls． The aim of the adjustment is to ensure that the
number of the steel balls on the estimated count plate is less than that of required． The steel balls will enter the
accurate count plate after estimated counting． The steel balls images are captured by the CCD camera which is set
above the accurate count plate． Then binarization is performed to get the binary images of the steel balls，and the
double gray threshold algorithm is adopted to calculate the number of the steel balls． The single ball compensating
device is used to compensate the number difference between the accurate counting and the number of being needed． The steel balls of abnormal size are recognized through comparing the distances between the center of two
neighboring steel balls in a defined district of the accurate count plate one by one． According to the experiments of
the prototype，the system can be used to count steel balls whose diameter are between 4． 763 mm and 25． 4 mm，
the accuracy of counting and the determine rate of the abnormal size steel balls of this system are all up to 100%．

In order to solve the problem of implicit opinion targets recognition in course comments，a method of implicit opinion targets identification based on text classification was proposed in this paper． First，the word vectors corresponding to the training text were obtained by word2vec model which produced the short text features． Then，TextCNN was used to extract high-level features by obtainning classification model by pooling K-max and putting it into softmax classifier． Finally，the trained classifier was used to classify the implicit opinion sentences，and the corresponding opinion targets of the implicit opinion sentences were obtained． The experiment results show that the attribute classification of implicit curriculum reviews based on the convolutional neural network，the accuracy rate of implicit opinion targets identification in curriculum reviews is 89． 9%，which meets the needs of implicit opinion sentence in curriculum reviews．

The purpose of this study is to improve the MOOC ( Massive Open Online Course) interactive teaching mode in the context of“Internet +”，improve the platform's functional framework，solve the problems existing in the operation of the platform，guide users' practice process，simplify the difficulty of resource download and cooperate with each other． Emphasis is laid on the server and browser of integrated B /S architecture，J2EE development environment，and the layered design principle of integrated architecture SSH ( Secure Shell ) architecture． The platform is divided into four layers，namely，controller layer，data layer，user view layer and business logic layer，so as to achieve a significant improvement in development efficiency． This research is based on the existing MOOC platform to carry out a more in-depth analysis，further improve its shortcomings，enhance its interaction and practicability． The coupling of the two improves the teaching effect，to provide a platform for students to make pre-school learning efficient．

The PVD ( Pressure Vessel Design) problem is a classical practical problem to confirm the ability of algorithms． Solving the PVD problem by CGSA is difficult． Thus， CGSA9 + is proposed． First， the interdependence relationship among variables in PVD is analyzed，then the stochastic order of dimensions is identified． After that，the generation rule is updated based on the rule of CGSA and the current position of the global optimum． Finally，the solutions satisfing the constrained conditions is verified to evaluate the objective function． According to the experimental results，this algorithm solves PVD problem effectively and it performs well．

In order to improve the efficiency of military integrated logistics transport，a new logistics transport scheduling node planning model is designed． Firstly，the logistics requirements of each node are determined，and the digital simulation operation is carried out in the form of a subset of the requirements． By determining the subset of the requirements，the independent needs of current logistics are excluded． Demand clustering is simplified by corresponding matching sets to confirm the priority characteristics of current nodes． According to the clustering results，the current logistics nodes with no designated logistics order and the non-existent logistics constraint points are sorted，and the optimal logistics path is determined according to the constraint conditions to achieve logistics transport scheduling node planning． The experimental results show that the conflict rate of military integrated logistics is reduced by more than 20% and the unimpeded path length gap is 37% after the application of the node planning model，which proves that the model can effectively improve the logistics efficiency and has strong application advantages．

In order to improve the security performance of backup university data and reduce the loss caused by system damage or data loss in colleges and universities，a unified disaster recovery backup technology is proposed． Firstly，it is clear that data backup is the key of unified disaster recovery technology． The real-time detection of the state of the target system is carried out through heartbeat and checkpoint technology． Secondly，information is partitioned in the process of data backup in colleges and universities，and each part is supported by a thread． The transmission is completed according to the corresponding network protocol． While the accuracy and completeness of data are guaranteed，backup is greatly improved． Finally，through the quantitative processing of resources，data fault analysis and algorithm optimization，the performance of the unified disaster recovery backup technology is further improved to better meet the data security needs of colleges and universities． The test results show that the application of unified disaster recovery backup technology in the data security protection of colleges and universities can maximize the integrity and security of data，and has high application and promotion value in the data security of colleges and universities．

Online comment is an effective way for users to express their opinions or suggestions on commodities．Analysis of comments is the basis of developing personalized services and improving the performance of commodities． However，there are a lot of false comments in the network． Accurate identification of the key comments which represent users' real feelings has become a hot issue in the field of comment analysis． A key comment extraction method based on reply support was proposed． The proposed method starts from user's feedback behavior to comments，and focuses on two indicators: like comment and comment reply． By calculating the like comment rate and comment reply rate，the method obtains the comment's reply support，only extracts the comment with high reply support，thus eliminating a large number of false or useless comments，and improves the accuracy of key comment extraction． Finally，through the experimental comparison with the existing mainstream methods，it is verified that the proposed method has a high accuracy and recall rate．

In order to solve the problem of the lack of general detection equipment for on-board auxiliary power
supply of rail vehicles，we design a new automatic integrated test platform for evaluating the performance index of
on-board power supply in rail transit． The software of this platform is the main controller，and this paper
synthetically uses computer control，frequency conversion and voltage regulation，sensor measurement，network
communication and other technologies to simulate the working environment of electric vehicles，integrates the
existing hardware equipment，and builds a unified hardware testing platform． On the basis of this hardware
platform，the software measurement and control system is designed and developed to enhance the engineering
research ability of the laboratory for the life and reliability detection of on-board power supply，and provide test
data for the advanced maintenance of the corresponding products as the basis for the maintenance，which saves
the testing resources and improves the detection efficiency．

In order to ensure the integrity of massive data transmission under big data cloud storage，a massive
data transmission integrity control method based on attribute feature matching and association rules is proposed．
The high dimensional reorganization model of attribute features of massive data is constructed，and the feature
distribution of key information is obtained． The key information storage and distribution structure model of massive
data is designed，the key information feature mining and extraction of massive data is carried out by using
association rule method． The feature analysis and clustering processing of massive data are carried out by using
key information． And the key information storage nodes of massive data are designed by attribute feature matching
method． Using fuzzy subtraction clustering to store key information the storage node carries on the clustering
processing． In the massive data transmission，the data key information storage node transmission integrity degree
is used to realize the massive data transmission integrity control． The simulation results show that this method can
improve the efficiency of spatial utilization and the integrity of data transmission in cloud storage．

In order to improve the informatization level of network fault maintenance service in colleges and
universities，the author puts forward a network fault QR( Quick Response) code repair and maintenance system
based on WeChat public platform and student volunteer service． The system is supported by cloud computing
platform of Yunnan Normal University，with the help of new media and QR code technology of WeChat public
platform，and combined with the campus network fault maintenance business of student volunteers，achieving its
main functions． The experimental results show that the system runs correctly and smoothly，and can significantly
improve the maintenance level of the campus network equipment．

In order to solve the problems of low positioning accuracy and long time in traditional data positioning
methods，an accurate positioning method for audit information system with full coverage network data is
proposed． The nearest neighbor method is used to analyze the full coverage network data association of the audit
information system，the correlation quality function theory is introduced to transfer the multi-feature data obtained
by the sensor into the process of association decision，and the correlation measure is calculated． According to the
calculation results of the correlation measure，the target tracker is constructed through the improved KCF
algorithm，and the specific running state of the target is obtained by using the tracker． The tracking set and the
detection set of the target are obtained by using the data association algorithm． According to the data association
algorithm，the tracking set and the detection set of the target are obtained． The tracking results of weighted fusion
of three features can accurately obtain the specific position of each target data in the audit information system，
realizing the accurate positioning of the data． The simulation results show that the proposed method can quickly
and accurately obtain the specific location of the full coverage network data of the audit information system．

In the case of smoke generated in the early stage of the fire，the video captured by the fixed
monitoring probe is used to detect the fire smoke，and the frame difference method of the computer video
detection technology is used to obtain the area of the smoke movement． When the camera is not moving，two
consecutive images are obtained and the motion region in the video is detected． The video of the smoke is
segmented by the variational level set method，and the motion region is segmented from the background to detect
and analyze the smoke． Therefore，it is more accurate to judge the fire through the smoke，and the detection
result is more real-time，which improves the efficiency of the video command fire fighting work．