In order to comply with the current sustainable and intelligent development trend of the power industry, an efficient, accurate and reliable power information collection system is constructed applying all-fiber networks to intelligent electricity meter reading. In the local backbone communication network, plastic optical fiber is used as a signal transmission medium, while in the remote channel, quartz optical fiber is used for large- scale and long-distance data transmission. The corresponding optical transceiver modules are added to the measurement equipment and communication nodes, which can make the power meter reading system all-optical. The experimental results show that all-optical sultion will improve the successe rate of meter reading to 100% , reduce time consumption to 2. 1 s and power consumption to 50 mW. It will bring more intelligent value-added functions to the power industry, laying a solid technical foundation for the modernization of electricity and various energy industries.

In order to analyze the rules of vehicle trajectory and so as to improve the capacity of the weaving area of roundabout, the method of vehicle trajectory clustering at roundabout is studied based on space-time similarity coefficient. The space-time information of vehicle trajectory is analyzed for the specified vehicle trajectory, and the space-time similarity coefficient is calculated and clustered by spectral clustering. The trajectory clustering in a period of time in the intersection area is visualized. It is proved that the method can effectively reduce the data and dig out the hidden rules of the trajectory information, which provides a valuable reference for further decision-making work.

Because the LSD (Line Segment Detector) algorithm, when extracting continuous edges in an image, often has the problem of line segment discontinuity in results, we proposed an improved LSD algorithm based on information entropy. Firstly, the algorithm calculates the mutual information entropy between the processed image and the original image to determine the number of layers of gauss pyramid and the number of images in the layers, to build an adaptive Gaussian pyramid. Secondly, the improved otsu threshold is used to divide the image into different regions according to the gradient peak value of the image,and the corresponding gradient threshold is calculated to separate the image background. Finally, line segments are found according to the gradient angle and verified by helmholtz criterion. The simulation results show that the algorithm solves the problem of LSD algorithm in extracting the discontinuous line segments. Compared with other popular algorithms Hough Transform, PPHT(Progressive Probabilistic Hough Transform), LSWMS(Line Segment detection using Weighted Mean Shift), LSD, EDLines, more meaningful Line segments were extracted.

The traditional stability control technology of heterogeneous communication network is prone to signal distortion in the transmission process, and the stability control effect is not obvious. Therefore, a stability control technology for heterogeneous communication networks with delay constraints is studied. According to different types of electronic signals, different acquisition templates are used to process information. Pattern recognition,probability method and conversion algorithm are used to identify the output results of stability control. The work flow is simplified, and is divided into four steps: initial data check, free network establishment, displacement change rate establishment and fluctuation value control, effectiving improving the anti-interference ability of
heterogeneous communication network system. The experimental results show that the stability control technology can effectively improve the control efficiency and keep the heterogeneous communication network running in a stable state.

To quantify the degree of the information recoverability in the noisy channels, a concept of information fidelity is proposed. We reveal the intrinsic relationship between the information fidelity and the mutual information, and also recover the relationship between the information fidelity and the error probability of the information recovery methods. The theoretical analysis indicates that the maximum information fidelity can be used to characterize the optimal distribution of the transferred messages. Due to this characterization, we present the optimal information transfer problem based on the maximization of the information fidelity. Through the case of binary memoryless channel, we demonstrate the relationships among the maximum information fidelity, the channel capacity, and the optimal distribution of the transferred messages. The theoretical and numerical results imply that the calculation on the maximum information fidelity can be simpler than that on the channel capacity.

Aiming at the problem that the accuracy and stability of the control system are easily affected byenvironmental noise, equipment performance and other factors in the process of robot vision servo, an adaptive gain algorithm based on infinite norm optimization task sequence ( INOTS: Infinite Norm Optimization Task Sequence) is proposed. By dividing the real-time global task into different subtasks, the gain is automatically configured according to the characteristics of the subtask at different times. The simulation results indicate that compared to the fixed gain and the traditional adaptive gain adjustment algorithm, the proposed gain algorithm can improve the response speed of the system effectively, enhance the anti-interference ability of the control system, and improve the anti-interference ability of the control system.

The main problem to solve the robot visual servo control in unstructured environment is to obtain the interaction matrix. The common problem to solve the interaction matrix is the singularity of the pseudo-inverse of the interaction matrix. Aiming at this problem, a new image-based visual servoing control method is proposed, using the incremental extreme learning machine to solve the problem of pseudo-inverse approximation of the image Jacobian matrix. In order to improve the convergence speed of the system, a speed improvement controller with an adaptive factor is adopted. Finally, a six-degree-of-freedom manipulator simulation is used to verify the effectiveness and advantages of the proposed method. The algorithm improves the robustness of the system and avoids the problem of calculating the pseudo inverse of Jacobian matrix.

In order to resovel the management problems of batteries and electrical appliances existing in the use of RV(Recreational Vehicle) . A kind of power management system of RV with raspberry PI 3B + as the main controller is designed. We design two parts of vehicle battery monitoring module and electrical appliance monitoring module. The battery monitoring module uses the battery-specific monitoring chip DS2438 to detect and manage the battery temperature, voltage and other on-board battery information, and to complete the status display and fault alarm prompt of single battery. The electrical monitoring module uses RN8209 chip to detect the electrical power of the electric appliance for the RV and to carry out intelligent management of the electric appliance through the main controller in time. The test results show that the system can accurately measure the relevant information of batteries and electrical appliances, and has certain practicability. And in view of the difficulty of traditional SOC ( State Of Charge) prediction, a modified AH
integration method is proposed, which fully considers the problem of battery capacity difference in actual use. The Matlab simulation results show that the method has high estimation precision and can be used for SOC estimation strategy.

When traditional denoising methods are used to deal with the image interfered by high intensity noise,the noise can not be removed effectively and secondary pollution is easily introduced in the repair process.In order to solve this problem, a kind of marginal figure oriented non-local means filtering algorithm is proposed.First of all, the second order difference edge information is obtained, and similar blocks within the scope of the non-local is searched. Then, the edge guided image and noise image are both used to generate filter weights, so that the edge information oriented non-local collaborative filtering framework is constructed. Compared with the local linear filtering method represented by the traditional filtering, the proposed algorithm can explore the edge information of the image and use a new non-local collaborative filtering framework for image denoising enhancing
the edge repair ability under the environment of high intensity noise interference. Experimental results show that under the condition of high intensity noise pollution, the improved image gets higher measurement index, and has better visual effect.

In view of the problem of data being attacked and stolen when managing medical sensitive information, an intelligent encryption algorithm of medical sensitive information based on quantum computing is proposed. The key construction of quantum encryption for medical sensitive information is designed by considering the mixed entangled state, and the protocol subspace matrix constructed inside the key is analyzed.By calculating the neighborhood distribution function of information, the separation matrix quantum coding is constructed to provide processing data for key rearrangement. Considering the interference of quantum entangled states and their additional states, the effective key encryption is carried out, and finally the intelligent encryption
of medical sensitive information is realized. The experimental results show that this method has good anti-attack ability, high transmission efficiency of encrypted information, and good overall performance.

In order to accurately identify the complete action sequence of video target individuals and enhance the accuracy of video behavior recognition, a deep network time grouping behavior recognition method based on over-limit learning machine is proposed. First, the number of states of the behavior recognition model is delermined according to the number of key human behavior gestures, establish the multi-scale structure association of human motion behavior, and the different scale characteristics of motion trajectories and edge contour wavelet moments are introduced into the behavior model to obtain general characteristics of human motion behavior. Using the video grouping sparse sampling method, the video is divided into equal duration groups, and the standard backpropagation method is used to optimize the model parameters, to realize low-cost video-level time modeling and to ensure the integrity of the modeling process information. Finally according to the hidden layer activation function output and corresponding output layer weight coefficients, sensitivity analytical formula is obtained, hidden nodes are sorted according to sensitivity parameters, minor nodes are deleted, and accurate recognition of deep network time grouping behavior are realized. The results of simulation experiments show that the method has a high level of recognition accuracy, is less time-consuming, and has strong robustness.

According to the construction requirements of SME (Small and Medium Enterprises) information service platform, in order to solve the problems of decentralized management and low resource utilization caused by traditional SME information system shaft deployment, the overall framework of SME service platform is proposed, and the experimental deployment of SME cloud platform is performed based on OpenStack open source cloud framework. Finally, a preliminary study on the high availability technology of SME service cloud platform. The results show that the platform framework based on basic resource layer, data middle layer, business middle layer and application layer solves the problem of information isolated island caused by traditional information system shaft deployment. And the prototype system based on OpenStack framework verifies that OpenStack is more suitable for building private cloud platform for small and medium-sized enterprises.

In order to solve the problem that the effect of sorting task aggregation method is not clear, under the concept of group intelligence, a comparative study is performed on the aggregation method under the condition of the grouping effect of the sorting task problem, attempting to find the sorting aggregation method. A concept of evaluating group wisdom-group wisdom validity is proposed. Empirical research methods are used to rebuild the sorting task of the group experiment. Based on the sorting data of the experimental test, seven different sorting aggregation methods are used to aggregate all the individual sorting of the sorting task problem to obtain the corresponding group sorting. The deviation between the seven kinds of group ranking and the real ranking is measured and the performance of different aggregation methods according to the size of the group wisdom validity is judged. The experimental results show that, among the clustering methods of reconstruction and ranking tasks based on the validity of group intelligence, the mode of the majority method performs best, followed by the median method and the K-Y method.

In order to arouse students' enthusiasm for learning hepatobiliary surgery, deepen their understanding of liver imaging and surgical treatment, and improve their learning efficiency and performance. A novel teaching method combining 3D visualization technology with traditional teaching is proposed. 90 clinical five-year students from Jilin University Medical College were selected for a randomized controlled study. The control group adopted traditional teaching methods of liver surgery, while the experimental group adopted 3D visualization technology combined with traditional teaching methods. After the study, the two groups of students were given a standardized test, and the teaching effect was evaluated by questionnaires. The results show that the comprehensive test scores and excellent rate of the experimental group were higher than the control group, and the teaching effect of the experimental group was better than the control group.

In order to realize the automatic identification of bone marrow blood cells, bone marrow erythroid and granulocyte data sets are constructed, and a CellNet network model is proposed based on deep learning semantic segmentation technology. The model increases the depth of the network by adding a residual module, uses a convolution residual block to make the network model easier to train, and combines the U-Net clipping operation to provide more refined features for segmentation. The experimental results show that the correct recognition rate of this model for bone marrow erythroid cells and granulocytes reaches 93. 65% and 95. 25% , respectively, which provides a method for automatic identification of bone marrow blood cells.

To improve the current stage of aquaculture, underwater environment monitoring, water quality testing, and other fields of personnel operation difficulties, low efficiency, an observation type underwater robot with water sample collection function is designed. It is equipped with sensors and water sample collection devices to collect environmental water information. Firstly, the experiment of water sample collection is carried out to verify its movement ability and water sample collection ability. Then aiming at the images collected from natural waters, the dynamic threshold white balance method is used to eliminate the influence of light source intensity and color deviation. Finally, we use three algorithms: contrast stretch, histogram equalization and contrast
limited adaptive histogram equalization (CLAHE: Contrast Limited Adaptive Histogram Equalization) to enhance the image, and choose contrast, information entropy, and average gradient to evaluate the image quality. The experimental results show that the design has the characteristics of flexible operation and strong interaction, and the CLAHE method improves the details of the underwater image. The design provides a new way for scientific research experiments, underwater environment detection, water quality detection, environmental protection, etc.

In order to solve the problem of waste classification in university campus, a method of garbage image classification based on convolution neural network and normalization technology is proposed. Without complex processing of the input image, the network model can extract image features according to the
algorithm. By cooperating group normalization and each layer of the network model, the shortcomings of the traditional classification algorithm can be overcome and the garbage image can be classified. The final recognition has a high accuracy rate, and can identify unrecyclable garbage and recyclable garbage.