Aiming at the problems of traditional TV remote control，such as the difficulty of searching favorite program in the mass program list and the complicated control of the fixed key，an action planning framework is constructed for the television remote control． In the framework， based on the action language， an universal inference rule that integrates user's needs and program information is established， and the intelligent search of relevant contents of the program is realized by the knowledge representation and logical reasoning function of action languages，and can use behavior add users on filling in the exclusive rules of
language． The application example shows that the television programmer's remote control can intelligently understand the needs of users，so that users interact through the voice channel names，program content to watch favorite programs， to define their own program search restrictions， achieving the definable and intelligent searching aim．

Time picking is a crucial step in microseismic data processing，the picking results have great influence on hypocenter location． Especially when the SNＲ ( Signal-To-Noise Ｒatio) is low，it is difficult to obtain arrival times accurately with conventional approaches． We propose a new time picking approach based on the AIC( Akaike Information Criterion) and Shearlet transform named the Shearlet-AIC to solve the problems mentioned above． We divide microseismic data into several scales according to different statistical characteristic between signals and noise to obtain the feature of different frequency domain． We can acquire the minimum values that represent the picking results of frequency domain by using the Shearlet-AIC． To verify the reliability of the method，we experiment it on both synthetic and field datasets． The results show that our method can precisely pick the arrival times even when the SNＲ of data is as low as -13dB and the accuracy is superior to some existing method．

 Some denoising method can not attenuate the noise in desert seismic signal because it has both low frequency noise and Gauss white noise in desert seismic signal． CSD( Compound Sparse Denoising) method is proposed to attenuate the low frequency noise and Gauss white noise in desert seismic signal simultaneously． In the compound sparse denoising，square loss is used as the fidelity constraint of the loss function，the L1 norm of the first derivative of signal and the signal itself are used as sparse constraint of the loss function． The Gauss white noise in the signal can be removed by minimizing the loss function． In order to remove the low frequency noise at the same time，compound sparse denoising combines the low pass filter so that the low frequency noise and the Gauss white noise can be removed． Tests on both synthetic records and real desert seismic signal indicate that compound sparse denoising can attenuate both low frequency noise and Gauss white noise in desert seismic signal． The composite sparse denoising is better than the bandpass filter in amplitude preserving，and the bandpass filter will produce more serious distortion after denoising，while the distortion of hte composite sparse denoising is smaller．

To solve the problem of lack of precision calibration target for survey vessel at sea，a research on the technology of course calibration identificationfor inertial navigation equipment by electro-optical theodolite from three aspects: selecting star，leading to star and calculating record data is conducted ． Combined with existing conditions and technologies，an algorithm for filtering and extrapolating the data of ship shaking with sine wave fitting method was proposed． This algorithm obtains the sine curve fitting parameters by the least square method after the parabolic difference subdivision of the ship shaking data． During the sea with “Yuan Wang”survey vessel，the method described was verified． The results show that the method can improve the isolation of ship shaking，the ability of tracking space targets and the accuracy and consistency while calculating the course error．

To solve the problem of traffic accidents due to fatigue driving，we use video image processing related characteristics of fatigue analysis technology，use Adaboost face detection algorithm based on training precise positioning driver face and eye area，real-time acquisition eyes area binarization，a judging threshold value comparison method is adopted to improve the blink of an eye，and extract the AECS( Average Eyelid Closing Speed) eyelid fatigue parameters， the PEＲCLOS ( Percent of Eyelid Closure over the Pupil Time ) ， a comprehensive state of fatigue analysis，ultimately determine whether fatigue driving． The experiment results show that the precision of the face and eye detection has greatly improved，the designed software can detect the
driver's fatigue state by real time and avoid fatigue driving．

General precise electrochemical battery model of Lithium-ion battery，Quasi-two-dimensional model，has two major numerical problems: excessive state quantities and high computational complexity，making it difficult to be applied in real time conditions． Aiming at the problems of quasi-two-dimensional model，a one-dimensional model is proposed． And the Laplace transform is used to transform the model into the frequency domain． The radial diffusion of solid-state lithium-ions is represented as a non-dimensional admittance form，which solves the problem of excessive state quantity by reducing the dimension． For the transcendental equation，the Padé approximation is used to reduce the computational complexity． The proposed model's complicity is analyzed and the model was verified by different discharging currents． The concentration distribution and terminal voltage of the proposed model and the exact model are compared． The experimental results show that a voltage error is within 40 mV，reducing the computational complexityunder and ensuring the accuracy of the model．

In order to reconstruct the three-dimensional shape of the rotating body，a single image is proposed to achieve the extraction of the rotating body bus and the symmetrical axis． This method is to obtain two-dimensional image by using industrial camera． In the image preprocessing，the image shooting causes tilt，and we gives a geometric correction method，bus information extraction of rotating body using the contour feature of target area，combined with the symmetry axis of symmetry around the rotating body contour curve to obtain the target object． With the aim of ceramic vase，the extraction of the bus and symmetry axis of the rotating body is finally realized，and the three-dimensional shape of the target is rebuilt． The experiment proves the effectiveness of the proposed method．

In order to improve the accuracy of the cam grinding process，the lift error of cam grinding is studied．Genetic algorithm is used to correct the cam lift error． The measured cam lift value is adjusted by the software provided by the first machine tool plant in C city． The cam lift error and the theoretical cam lift value are taken as inputs，and the two order difference values are obtained by wavelet transform． The smooth two order difference of cam lift value is taken as the objective，and the genetic algorithm is used to correct the cam lift error． The simulation experiments show that this method can obtain a smooth cam lift values．

 In order to improve the speed of ContourExtractor2DimageFilter algorithm and achieve real-time image contour extraction，FPGA ( Field-Programmable Gate Array) is chosen as the main processing chip to study how to implement the hardware part of ContourExtractor2DimageFilter algorithm based on Xilinx Spartan-3 series FPGA design． In order to improve the speed of the algorithm and the real-time image contour extraction，more effective computer systems are used to interpret images，to achieve a similar human vision system，to understand the external world．

There are few methods to restrain noise in the traditional method of measuring wind． In order to better suppress the noise problem，a new method of wind measurement is proposed，which bases on the MVDＲ( Minimum Variance Distortioniess Ｒesponse) beamforming algorithm and combines with a special arc array．Ｒedundant spatial sampling information is obtained by using a four element ultrasonic sensor arc array，and the information received by the array is processed by the MVDＲ beam-forming algorithm to estimate the wind speed and direction information that to be measured． The simulation results show that the estimation of wind speed 0 ～ 60 m /s and wind direction angle 0 ～ 360° can be easily realized． The method is superior to the noise resistance and can be used to measure the wind speed and wind direction angle accurately．

Because the present J-V measurement systems for solar cells were complex， expensive and inconvenient to use，a new system which can auto-calibrate the xenon lamp irradiance and automatically measure several devices simultaneously was developed． The system consists of a computer，a testing program，a Keithley 2 400 source meter，a 500 W xenon lamp and its power supply，an automatic switching and measuring circuit for multiple devices，a xenon lamp operating time measuring circuit，a photocurrent measuring circuit composed of a reference photodiode and an irradiance calibration circuit． PC testing software was written in Visual Basic． The light power of the Xenon lamp was monitored using the photocurrent of a reference photodiode illuminated by the
lamp and calibrated to standard 100 mW /cm2 AM 1． 5 solar irradiance before every J-V measurement via modulating the output voltage of the Xenon lamp power supply using a microprocessor． The photocurrent of the custom solar cells under different bias voltage was measured using a Keithley 2400 source meter and a multi-sample automatic measurement circuit． The J-V curves were plotted and the device parameters such as efficiency were calculated by the testing software． The system proposed is cost-effective，highly automatic and fulfills the requirements of solar cells J-V testing．

As people get deeper and deeper into the scene audio research，the existing analytical methods can not satisfy the needs of people． The analysis method based on time-frequency domain characteristics can solve this problem well，by extracting the spectrogram of scene audio，the acoustic events included in the analysis signal can be fully retained，and it make the acoustic characteristics more intuitive． The spectrogram of scene audio contains rich texture information，by setting different length of the window，we can get the wide-band spectrogram and narrow-band spectrogram of the scene audio separately． The comparison experiment shows that the narrow-band spectrogram can better reflect the trend，continuity and distribution characteristics of the acoustic events in the analysis signal． It has better characterization than the wide-band spectrogram． Therefore，if the time-frequency domain characteristics analysis method is used for scene audio，it is best to use narrow-band spectrogram．

 In order to solve the problem that the data of the traditional surveying and mapping industry can not be real-time mapping，real-time imaging，and data and image can not be quickly integrated modeling，a panoramic laser system based on the Web platform which integrates Ｒaspberry Pi and embedded microprocessors and intelligent mobile platform is designed． The system uses laser radar as a data acquisition tool for the mapping area，uses NＲF24L01 wireless transceiver and Wi-Fi module for data transmission and reception，uses Ｒaspberry Pi and STM32 microprocessor for data analysis and processing，and the user can use Web client to query images and data． The system uses the intelligent mobile platform to measure the data，realizing the data storage on the network platform，and realizing the rapid modeling and imaging of the mapping data． The system achieves stable operation of low power consumption under harsh environmental conditions such as bright light，no light，mountains and tunnels，and completes the surveying and mapping area point spacing，partial area，full area measurement，historical data reconstruction and panoramic image rolling display． The author designed the panoramic laser system combined with software algorithms and user-friendly interface to solve the problem that in the traditional surveying and mapping industry orientation，ranging，height is not convenient enough，the imageand data can not be real-time integration，greatly reducing Installed capacity and mapping cycle．

In order to solve the problems of traditional constant temperature control in industrial and agricultural applications，such as low automation level，low precision and poor stability，an automatic control system based on novel fuzzy-PID ( Proportion Integration Differentiation) control theory is designed． STM32 microcontroller is used to collect temperature and control，matrix keyboard and LCD ( Liquid Crystal Display) screen are used as human-computer interaction equipment，the novel fuzzy-PID control algorithm is used to construct the fuzzy-PID controller，the parameter values of proportion，integration，differentiation are processed by iterative algorithm． The semiconductor chiller is heated by the corresponding PWM ( Pulse Width Modulation) high level duty ratio output current based on the control quantity． The experimental results show that the system can quickly collect the temperature data of the system and control the system in real time，stabilize the temperature of the system at the expected temperature，the absolute error of the system is 0． 1 ℃ in theory，the actual measurement error is 0． 5 ℃． The system could achieve a better automatic detection and control of temperature with reliable performance，and solve the above problems．

 To realize supervising and adjusting the concentration of carbon monoxide of underground garage，Concentration detection of carbon monoxide in the underground garage，alarm and ventilation control are studied． An underground garage ventilation control system based on TMS320F2812 is designed． It consisted of four parts: centralized controller，data collection and actuators，communication module and site control device．Micro -controller is used to control the work site． It was used to adjust all induced draft fan unified centralized control mode and parameters． Data acquisition and actuators included CO ( Carbon Monoxide) concentration sensor and microprocessor． It was responsible for receiving the host computer's various operations control commands and parameter settings to achieve the concentration of CO detection and dynamic display． Site control device consisted of relays，induction fan，alarm and 12864 display components． The experimental results show that a good energy-saving effect can be achieved by intelligently controlling the inducing jet
exhaust． This article provides a theoretical basis for the development of ventilation automation in underground garages．

 To extract diversity image features，an improved convolution neural network structure is proposed．The network introduced a multi-layer recursive neural network． Firstly，the network learned the shallow features of input images with convolution neural network，and then learned high-level features through the convolution neural network and multi-layer recursive neural network at the same time． Finally input two kinds of high-level features fusion into the classifier． This paper used the theory of transfer learning to solve the problem of the small sample training set data，and applied the improved convolution neural network structure to the oil pipeline steel grade identification． The experiments explore the influence of recursive neural network number and convolution kernels number separately，and make sure the final structure． The results show that the improved convolution neural network gets a low error rate of 3% ．

Because the accuracy does not met the practical requirements， a shallow residual network is proposed． With the characteristics of the network architechture，the network can combine features from different levels and calculate． The experiment shows that the shallow residual network has effectively improved the accuracy of the algorithms．

In order to solve the problem that obtaining effective information is difficult for users due to serious network information overload，a new method of network information recommendation is proposed． Firstly，a topic model for each user is built，the establishment of theme model which combines Newton cooling rate is to balance the effect of the time factor on the user interest． Then，collaborative filtering algorithm and content-based recommendation algorithm are used to meet the user's needs for information diversity and personalization． The experimental results show that the algorithm can effectively improve the accuracy and recall of recommendation，and the prediction of user preference effect is good． It is an effective recommendation method．

In order to solve the problem of discrete knowledge，poor systematization and poor applicability of knowledge model in online learning platform． We propose a knowledge model based on knowledge map，and then use the FP-Growth data mining algorithm to mining association rules to update knowledge model．Practice shows that the knowledge map can effectively improve the systematicness of the knowledge model． The FP-Growth algorithm fully considers the influence of student factors on the knowledge model and improves the applicability of the model．

In order to improve the performance of multi processor system，the calculation process，inter process communication time，the speed of the system，the effect of overhead in multiprocessor systems，and the multiprocessor system overhead acceleration coefficient are presented to improve the speedup effect，the order of parallelization，the maximum speedup of seven factors，providing an effective method for the optimization of multiprocessor system．