A study is conducted into the process of getting customer demands during the process of product design and development finding the way of avoiding the missing information during investigating customer demands. Based on the theory of affair-element in Extenics, a formal research is made on the front-end process of product development from user "needs", that is, the sources of the user demands. The methods of acquiring user needs and setting up the affair-element models of user needs, the importance degree analysis methods and the calculation model of user needs, the method that user need models to map into the product function models, and so on. Through a formal research of user needs and products' functions, a good solution can be derived to the problem for the product development of information on the first floor, which will contribute to the product development team in developing new products meeting user demands.

It is believed that the variation is useful to improve the learning of the concepts. Extenics explores the transform and reasoning of a problem. The method of variation based on Extenics was put forward. This process of variation based on Extenics can be shown in three respects as follows: modeling of basic-elements, situation variation of basic-elements, and application variation of basic-elements. An example was introduced and the advice was put forward.

If the initial iterative performance index function is a positive semi-definite function, then the value iteration of adaptive dynamic programming will converge to the optimal. This is the convergence condition of value-iteration based adaptive dynamic programming. Based on the condition, the initializing and updating methods for iterative performance index function is studied and a cooperative optimization algorithm based on adaptive dynamic programming is proposed. The simulation results show that the proposed algorithm can rapidly reduce the iteration residuals and greatly improve the convergence rate of adaptive dynamic programming.

Convolutional neural network (CNN) is a deep learning model with strong expression and classification ability, and is currently widely used, but there are a variety of specific algorithms. In this paper, the realization of deep learning algorithm based on convolutional neural network CNN, including the function of convolution kernel, the role of pooling, the selection of activation function and the training process are discussed. And one example is explained, which facilitates the mastery of CNN. Finally, the future research direction of convolution neural network is summarized and forecasted.

With the Internet technology and modern E-commerce becoming popular, the recommender system has been widely used, but two problems of most recommendation algorithms still remain, i.e. cold start and explanatory problem. Based on the HFT (Hidden Factors and Hidden Topics) model which combines the review and rating information, an improved HFT model is proposed. By adding the free vector in order to capture the review information not discussed in the HFT model, the two problems can be released and the model accuracy improved. At last, the two large datasets shows that the proposed model is better than the HFT model in accuracy, which can largely benefit the use of review information.

The robot's self-localization is the key to realize navigation and intelligent. To deal with the accuracy of location problem of the flying robot in indoor environment, an improved visual odometry/SINS integrated localization algorithm based on BA is proposed. First, the proposed method combines the calculation of visual information by direct method with the angular velocity and acceleration information of inertial unit, and iterate using extended Kalman filter method, which enhances the robustness of the visual odometry. The estimation preciseness of feature points' depth information is improved with inverse depth method. Then bundle adjustment is used for local optimization. The experiment results show that the proposed algorithm has effectively improved the preciseness of robot-pose estimation.

FREAK algorithm has the defects of not having scale invariance and single feature point matching strategy, and being prone to unsatisfactory results. Based on SIFT and RANSAC algorithm, an improved FREAK algorithm is proposed: SFREAK (SIFT and FREAK). First of all, in the generation of Gauss differential Pyramid image, the feature points are detected with scale invariance; then the feature points are described with FREAK descriptor, obtaining binary descriptor; finally, in the process of feature points matching using Hamming distance matching for coarse matching, the matching points are purified with RANSAC algorithm, and the feature points of two images are matched. The experimental results show that the proposed algorithm can effectively solve the FREAK not having scale invariance in image scaling, and SFREAK algorithm for feature point matching accuracy rate reached 95.7%, increased by 61.9% compared with FREAK. Therefore, compared with the traditional SIFT algorithm and FREAK algorithm, the improved algorithm shows better robustness.

In order to solve the ill-posing problem and poor effect of fixed regularization parameter in super-resolution image reconstruction, an adaptive regularization combining the study of sparse representation is proposed. By additional restrictions for compatibility of adjacent patches, a new L_1/2 non-convex optimization model is built. Reweighted L₂ Norm rather than L_p (0<p<1) Norm is applied into the adaptive algorithm for adjustment of regularization parameter. With the help of joint dictionary training method, some important features for improving the quality of reconstructed image are obtained. Experimental results show that the method has significant advantages in denoising and preserving edge details. It is showed that the proposed method not only makes the desired high-resolution images visually clearer, but it also outperforms some traditional methods in both the value of peak signal to noise ratio and structural similarity.

A new algorithm is designed for splicing aperture in image mosaic. This new algorithm is proposed to solve the horizontal direction and the vertical direction aperture at the same time which the gradually fading image fusion algorithm cannot accomplish. The weight function is designed on the closed disk according to a continuous map on the overlap region of the image to the closed circular plate. Then the algorithm is simulated. According to the simulation results, the algorithm can eliminate the gap in the process of image mosaic and achieve good visual effect, and has significant practical value in practical application.

Aspect-Oriented Programming (AOP) can effectively solve the code-tangling and code-scattering caused by crosscutting attributes; however, much work is presented on the programming and the implementation phase, and little has addressed the early model stages. A research is conducted on the requirements engineering phase of software by using AOP technology. First in the requirements analysis phase, the functional requirements and non-functional requirements are identified by the separation of concerns, and the functional requirements with components and the non-functional requirements with aspects are realized. Then the components and aspects are represented by combining with the formal language Aspect-Z which has the characteristics of accurate description. As actions accompany conflicts while two or more aspects affect the same joint point synchronously, a method is proposed to solve by defining aspects' priority level. In the end, the properties and attributes of the described requirements are deduced by the theorem proving method, so as to achieve the purpose of formal verification of Aspect-Z specifications. Finally, an application example is given.

3D laser scanning technology has been widely used in various disciplines, but current point cloud processing software lacks efficient algorithms and user interaction, causing large time and effort consumption. This research aims to develop a new point cloud processing software which includes data importing, filtering, segmentation and modeling based on PCL and Qt. It provides key parameters setting for each typical algorithm and user friendly interface to produce optimal results for each module.

With regard to the characteristics of the unit commitment problem, a hybrid crisscross optimization algorithm (CSO) is proposed, discrete CSO and continuous CSO are used for the unit status scheduling and power economic dispatch, respectively. In the process of evolution, horizontal crossover performs crossover on the same dimension of different particles to guarantee the global convergence ability, and vertical crossover performs crossover on different dimensions of the same particle to overcome the local optimal convergence, and the perfect combination of the two crossovers by a competitive algorithm not only speeds up the convergence rate but also ensures the ability of the global constraints. The heuristic strategy is introduced to modify the constraint conditions, such as the minimum up and down time and the power balance, which can effectively improve the accuracy of the solution. The feasibility and superiority of the proposed algorithm are verified by the simulation of two examples and the comparison with other algorithms.

Based on the advantage of GPS (Global Positioning System)-RTK (Real-time kinematic) rapid positioning technique, a GPS-RTK-assisted close-range photogrammetric bundle adjustment method is presented. On the basis of the conventional photogrammetric adjustment model, GPS-RTK observation equations are added. According to the relative offset between the GPS observation center and the camera projecting center, GPS-RTK observations are given appropriate weights and a rigorous adjustment is carried out according to the least-squares principle. The experiment results show that the parameters solving process is stable and the planar location precision can be 0.02 m, and the new adjustment method can reduce the accuracy requirement of the initial parameter value.

Index error of vertical circle and 2C value (two times of collimation axis error) are two important indicators which affect the precision of total station's angle measurement. A South NTS-312L total station with index error of vertical circle overrun and a South NTS-312L total station with 2C overrun are corrected by the calibrator stand with collimators. The experiments show that the index error of vertical circle of the former is reduced from 24" to 6" and 2C of the latter from 19" to -6" respectively after calibrating, the calibrating results according with Verification Regulation of Electronic Tachometer Total Station JJG100-2003. In order to check the calibration results further, both total stations are tested in a field survey, and the results show that the vertical index of the former changes between -3" to 6" range, and 2C value of the latter varies in -6" to 5" range, these numbers basically in agreement with the calibration experiment.

With the promotion of more and more stepper motor drivers on market, it is convenient to control and use stepper motors. In some cases the stepper motor needs running at high speed, and in order to prevent the loss of step, overshoot and stall, its acceleration and deceleration process need to be rationally designed. The theory of the acceleration process of stepper motor is studied based on the physical properties of the stepper motor and then the stepper motor acceleration curve generation tool software designed. This tool software can generate acceleration curve data automatically according to the stepper motor parameters and acceleration process demand parameters, and can be easily exported to .C and .H file for microcontrollers program debugging and using, and finally through the microcontroller control platform its effectiveness and practicability is verified.

BaTi_1-xSn_xO₃(x=0.02, 0.10, 0.15, 0.25) ceramics were synthesized by the traditional solid state reaction method. XRD spectrum shows that all sample is tetragonal phase. With the increase of Sn⁴⁺/Ti⁴⁺ molar ratio, Curie temperature of all compositions will drop a little. When x value range 0.02 to 0.15, the impedance spectrum analysis shows that the sample resistance may be mainly ascribed to grain boundary. While for composition of x=0.25, the response from grain can not be ignored. Through analyzing the dielectric spectrum and impedance spectroscopy, the values of activation energy calculated from conduction were lower than that calculated from impedance measurements. The results suggest that the relaxation process of BaTi_1-xSn_xO₃ ceramics is dominated by a dipolar conduction. The conduction mechanism can be explained by thermal dynamic process of oxygen vacancy.

In order to systematically analyze the effect of setting exhaust heat recovery device on the overall energy consumption of the central air conditioning system, the central air conditioning system with air-cooled chiller as cold source is selected as study object. On the condition of meeting the requirement of the thermal comfort and indoor air quality, the energy flow mathematical models for this central air conditioning system with and without setting device of exhaust heat recovery are established through analyzing the energy flows of the four consecutive loops which extract energy from the conditioned spaces and reject it to the environment. Based on the HVAC design parameters of a commercial building in Guangzhou, the energy flow diagrams of this air conditioning system with and without exhaust heat recovery are drawn respectively, which can visualize the main differences of energy flow and energy consumption between different systems and provide useful messages for selection and energy saving potential of the system.