Aiming at the insufficient robustness problem of weak signal detection and modulation recognition in multi-sensor distributed reception systems, a new joint processing method based on deep learning is proposed. The proposed method adopts the distributed soft information fusion processing strategy where the signal detection and modulation recognition are integrated into a multi-variate hypothesis test problem. With the help of the excellent function approximation ability of DNN(Deep Neural Network), a method of joint pos terior probability solution and classification based on deep neural network DNN is proposed based on the analysis of network structure, objective function and network input and output. Finally, the performance of the proposed method is verified by simulation experiments, and compared with the existing methods. The results show that the proposed method can effectively fuse multiple sensor signals, and can significantly improve the classification accuracy with the increase of the number of receiving units. Compared to the existing confidence fusion methods based on equal weight combination, the proposed method has better performance, which is more obvious at low SNR(Signal-to-Noise Ratio) values, short signal lengths and large receiving units numbers.

The estimation error of the time difference of the frequency hopping signal in the spread spectrum communication network is too high, leading to a decrease in signal positioning and tracking ability, and signal propagation stagnation in multiple fields. In order to enhance the propagation efficiency of the frequency hopping signal in the spread spectrum communication network and improve the signal detection ability of the radiation source tracking and positioning system. A method for estimating the time difference of hopping signals in spread spectrum communication networks is proposed based on FRFT ( Fractional Fourier Transform ) and blind separation. The blind separation method is used to obtain two channels that meet the time difference estimation conditions of frequency hopping signals, namely flat fading channel and frequency selective fading channel. The two channel characteristics are estimated by FRFT. The time difference estimation of frequency hopping signals in spread spectrum communication networks is realized by combining the characteristics of flat fading channel and frequency selective fading channel with the maximum likelihood block detection algorithm. The experimental results show that the root mean square error of the proposed method is at a low level and the estimation success rate is at a high level whether in normal environment or noise interference environment.

It is difficult to detect the tampering evidence after the images with tampering operations such as splicing and synthesis are compressed. In order to achieve tamper detection of compressed images, a compressed image stitching and synthesis detection method based on camera calibration is proposed. The imaging law of compressed image is analyzed, and the spatial transformation matrix is used as the detection evidence. Based on the camera calibration parameters, the homography matrix is estimated. Based on the external parameters of the camera, the authenticity of the image is determined by rotating the spatial transformation matrix of the image, and the detection of stitched synthetic compressed image is completed. The experimental results demonstrate that this method can detect images that have been spliced, synthesized, and tampered with, with a detection accuracy of 75% .

A GA(Genetic Algorithm) based C-IoT(Cognitive Internet of Things) power adaptive PIS( Partial Interference Steering) algorithm is proposed for the interference management problem in C-IoT(Cognitive Internet of Things) systems under the concurrent spectrum access model. The algorithm can improve the spectrum fficiency of the system while ensuring the quality of service for both the PU ( Primary User) and the CU (Cognitive User). The simulation results show that the algorithm can converge quickly in seeking the optimal spectral efficiency of the system and calculate the optimal transmitting power of the PU and CU desired signals. In the scenario where the relative positions of the primary transmitter, PU and CU are determined, the optimal spatial distribution of the cognitive transmitters with access to the authorized spectrum can be solved based on the average degree of constraint violation D_{cv_ave} by the users.

Aiming at the problem of inconsistencies between the voltages of the two capacitors on the DC(Direct Current) side of the midpoint clamping (Neutral-Point-Clamped) three-level inverter and the traditional SVPWM (Space Vector Pulse Width Modulation) can solve the problem of inconsistency of the two capacitor voltages within a certain modulation system, but the midpoint potential cannot be balanced under a large modulation ratio, the control method is optimized on the basis of the traditional SVPWM and VSVPWM ( Virtual Space Vector Pulse Width Modulation). The method is based on the flow direction of current, different positive and negative small vectors using different size of balance factor, is in accordance with the midpoint potential difference value, and introduces voltage adjustment coefficient. On this basis, it is combined with the beat-free control. This closed-loop control strategy adjusts the output waveform by using the midpoint potential difference and the inverter three-phase current output value as the feedback to suppress the midpoint potential. Simulation results show that the proposed method can still maintain the balance of the midpoint potential on the DC side when the three-level inverter modulation system is high, which proves the correctness and effectiveness of the control strategy.

In order to solve the problem of the unreasonable weight distribution of the evaluation index system and the uncertainty of air combat data in the evaluation process, game theory and interval cloud models are proposed for effectiveness evaluation of incomplete air combat decision-making. Aiming at the attack and defense decision of air combat, an evaluation index system is constructed, and the subjective and objective weights are reasonably adjusted based on game theory to obtain the comprehensive weight value of the index. Then, for the randomness and ambiguity in the evaluation, the interval cloud model method is studied, and the effectiveness of incomplete air combat decision-making is determined by the interval cloud generator. Finally, the feasibility and effectiveness of the proposed method are verified by simulation. It provides technical support for solving the problem of evaluating the effectiveness of air combat decision-making under incomplete information.

In the deep Reinforcement Learning, the PPO ( Proximal Policy Optimization) performs very well in many experimental tasks. However, KL(Kullback-Leibler) -PPO with adaptive KL divergence affects the update efficiency of KL-PPO strategy because of its asymmetry. In order to solve the negative impact of this asymmetry, Proximal Policy Optimization algorithm based on CIM( Correntropy Induced Metric) is proposed characterize the difference between the old and new strategies, update the policies more accurately, and then the experimental test of OpenAI gym shows that compared with the mainstream near end strategy optimization algorithms clip PPO and KL PPO, the proposed algorithm can obtain more than 50% reward, and the convergence speed is accelerated by about 500 ~ 1 100 episodes in different environments. And it also has good robustness.

In order to solve the problems of high packet loss rate, long link delay and low utilization in traditional wired hybrid network, a method of congestion optimization is proposed. Based on the detection results of one-way delay and available bandwidth, Markov model is used to predict the congestion status of wired hybrid network transmission, which takes resource consumption and link utilization as the target of optimization control. When the path of optimal control of the target meets the requirements, the congestion optimization control of data transmission of wired hybrid network is completed. The experimental results show that the packet loss rate of the proposed method is within 4% , the link delay is controlled within 0. 05 s, and the packet loss rate is low, the link delay is short and the link utilization is high.

The rock image recognition algorithm model based on traditional deep learning is cumbersome and requires certain computing power when it is applied to mobile terminals, so it is difficult to realize real-time and accurate identification of rock types. Based on the ShuffleNetV2 network, we insert the ECA (Efficient Channel Attention) module of the channel connection attention mechanism, use the Mish activation function to replace the ReLU activation function, and introduce the depthwise separable convolution in the lightweight network components. Experiments are performed on rock images with this method. Experiments show that the recognition accuracy of the algorithm reaches 94. 74% . The improved algorithm structure is not complex and maintains the characteristics of lightweight, which lays a foundation for its application in limited resource environments such as mobile terminals.

Information privacy of database is vulnerable to illegal attacks. It has low security and less storage space. In order to solve the problems an encryption storage algorithm database information privacy for based on chaotic mapping is proposed. The chaotic sequence is generated by disturbing the control parameters of Logistic chaotic mapping, and the sequence is mixed with the chaotic sequence generated by other systems to generate a new chaotic sequence. Then, the non-linear transformation is carried out through the dynamic coding algorithm. The output sequence is used as the database plaintext key to encrypt the plaintext. The index field for database encryption information query is constructed, compressed by hash function, and stored in the database together with the encrypted information realizing the information privacy encrypted storage of searchable chaotic mapping database. The experimental results show that the proposed method can shorten the time of encryption and decryption, reduce the space occupation and reduce the energy consumption.

In order to solve the problems of management, coordination and association of cross regional scientific and technological resource sharing and collaborative innovation, practical application and theoretical research of scientific and technological resource platform are carried out to build an ecological digital display system of scientific and technological innovation. First, the definition, composition and index system of scientific and technological innovation elements are constructed, and innovation map, innovation ability, innovation subject, innovation carrier are proposed. The digital display system with innovation resources and innovation environment is the core. Taking Jilin Province as an example, the index method is used to monitor the level of regional scientific and technological innovation, the text mining method is used to analyze the hot spots of scientific and technological policies, and the system architecture and function design are carried out based on the Vue + Django + MySQL technology architecture. Finally, the ecological digital display of scientific and technological data innovation in the region is realized, so that the system has data sharing, data association and intelligent service functions such as decision support. The application practice shows that the system enriches the visualization of scientific and technological data resources, improves the visualization of scientific and technological innovation ecology, and improves the expansibility, efficiency and performance of the system.

At present, the distribution room mainly relies on the traditional manual inspection method, which has the problems of high inspection cost and high false alarm rate of hidden trouble. Based on the knowledge map and intelligent cognitive technology, the basic theory and key technology of the knowledge base and advanced application of the operation inspection business of the distribution station building are expounded. The knowledge sharing management, knowledge intelligent search, knowledge intelligent recommendation, knowledge data label and fault intelligent reasoning are studied. The distribution station building knowledge base is established and applied in Shanxi power grid. The research results build a knowledge base for intelligent operation and inspection of power distribution station buildings, and carry out intelligent application research, realize knowledge empowerment such as intelligent analysis of fault hidden dangers, reduce the inspection cost, and improve the intelligent operation and inspection capability.

Because the traditional single image super-resolution reconstruction algorithm fails to make full use of the shallow feature information, ignores the spatial structure information in the visual target, is difficult to capture the dependence between the feature channel and the high-frequency feature information, and there are artifacts and edge blur in the reconstructed image, an image super-resolution reconstruction algorithm based on residual network and attention mechanism is proposed. The feature extraction part of the model combines the WDSR-B (Wider Activation Super-Resolution B) residual network to enhance the flow of feature information in the network, weights the feature parameters through the coordinate attention mechanism, and guides the network to better reconstruct high-frequency features and restore image details. The experimental results show that under quadruple image reconstruction, the PSNR(Peak Signal to Noise Ratio) on Set5 and Set14 test sets is 31. 00 dB and 28. 96 dB, and the SSIM( Structural Similarity) is 0. 893 and 0. 854. The reconstructed image performs better in detail and contour, which is better than other mainstream super resolution reconstruction algorithms.

To reduce the transportation cost in the VRP (Vehicle Routing Problem) of milk-run, a discrete seagull algorithm is proposed. Firstly, in the process of seagull migration, insert and reverse operations are used to update the seagull position to improve the algorithm's search speed. Secondly, swap and 3-opt operations are used to update the seagull position to improve the algorithm's local search capability. Finally, combined with simulated annealing algorithm, the phenomenon of landing on local optimum is prevented during the operation of the algorithm, the update strategy is redefined under the discrete vehicle routing problem. With the lowest total cost as the objective function, the corresponding mathematical model is constructed. Experiment results show that the algorithm is able to efficaciously deal with the vehicle routing problem of milk-run, the finding effect and solution quality are better than the standard seagull optimization algorithm, particle swarm algorithm, simulated annealing, gray wolf optimization, whale optimization algorithm, and moth-flame optimization.

The unique advantage of dynamic fuzzy logic programming language is that it can process dynamic fuzzy data, but the existing compilers are difficult to effectively parse dynamic fuzzy data. To solve this problem, a new dynamic fuzzy logic programming language compiler is designed by extending the structure of supervised command program and introducing the formal description of dynamic fuzziness. The example shows that the compiler can correctly parse dynamic fuzzy data. Furthermore, it can reduce the difficulty of dynamic fuzzy logic program debugging and improve the efficiency of dynamic fuzzy logic program development.

Aiming at the problem that the traditional distribution strategy has a difference in available capacity in the hybrid energy storage system, and the hybrid energy storage system will be out of service due to insufficient available capacity, a power distribution strategy using improved sparrow algorithm is proposed. The ratio of the effective storage capacity to the overall capacity in the system is the optimization objective. And the improved sparrow algorithm can better solve the power distribution problem between lithium batteries and super capacitors. Aiming at the characteristics of high power and low energy density of supercapacitors, and the problem of insufficient available capacity in practical work, a method using lithium-ion batteries to adjust the residual effective energy storage capacity of supercapacitors according to the transfer current is proposed. The controlled transfer current solution method ensures that the supercapacitor always maintains a certain effective energy storage capacity, thereby enhancing the continuous operation capability of the supercapacitor. Finally, the rapidity, stability and effectiveness of the strategy proposed are verified by simulation.

To inherit and protect the traditional blue calico patterns, it is necessary and challenging to protect blue calico in a digital manner. Due to the lack of blue calico pattern datasets with original manual characteristics, which limits the application of powerful deep learning technology on pattern recognition field like the recognition of Blue Calico patters, therefore a large-scale dataset called Blue-Calico pattern dataset is provided, which is the first publicly available benchmark for blue calico pattern recognition. This dataset contains about 50 216 blue calico patterns, covering 85 pattern classes, such as animals, plants, myths and legends. The construction of this dataset will concern the digital construction of blue calico, such as calico image retrieval and caption, and enable researchers to design and validate data-driven algorithms. On the basis of the new dataset, the experimental results of four state-of-the-art networks are provided as a baseline for future work.

In the face of problems such as low resolution and image blurring in drone inspection images, a super-resolution reconstruction method is proposed for drone inspection images using the theory of VDRCNN(Very Deep Residual Convolutional Neural Network). The algorithm model consists of a VDSR( Very Deep Network for Super-Resolution) and a residual structure. Based on the VDSR, the algorithm is improved by adding a residual structure to enhance convergence speed, while combining batch group normalization and Adam optimizer to achieve better reconstruction effects. On this basis, an electric power component detection dataset is constructed, and high-resolution reconstruction of blurred electric power component images is achieved by properly setting the network parameters. The experimental results show that the super-resolution method based on VDRCNN can reconstruct images with richer textures and more realistic visual effects, with improvements of 2. 95 dB and 3. 79% in peak signal-to-noise ratio and structural similarity respectively, compared to traditional detection methods. Therefore, the proposed VDRCNN-based super-resolution reconstruction method has certain potential application value in solving practical problems in power inspection.

A new data filtering and fusion algorithm are proposed for the problems of ineffective energy consumption and short network lifetime in oil and gas IoT( Internet of Things). This algorithm can adaptively judge the degree of data abnormality, filter and fusion data, avoiding redundant network information and excessive energy consumption. The algorithm adaptively determines the abnormality of the data by judging the deviation degree between the monitoring data and the normal data, performs intra-cluster filtering and inter-cluster fusion on the data. Compared with the traditional scheme, the proposed scheme can effectively improve the communication quality and energy consumption efficiency of oil and gas IoT.

Convolutional neural network model has the disadvantages of large volume, high computation and poor performance in small and resource limited embedded platform. The existing lightweight model can not take into account the detection speed and accuracy. The mainstream target detection algorithm Yolo v4 is selected to lighten the model, and the mobilenet network and depthwise deparable convolution are used in Yolo v4 model. The results show that compared with the original Yolo v4 model, the improved Yolo v4 model of different mobilenet networks can process an image about 19 ms faster on average, and the accuracy rate can reach more than 92% . The accuracy rate of the improved Yolo v4 model with mobilenet v3 as the backbone feature extraction network is 95. 13% , which is 2. 99% higher than of the original Yolo v4 model. The parameter of this model is about 1 / 6 of Yolo v4 model, and the model can process a patrol image 20 ms faster than the original Yolo v4 model. Insulator is an important part of transmission line, The identification of insulators in many images can help to analyze the operation of transmission lines.

In order to realize the high-precision application of digital images and reduce the influence of external light on visual imaging, a fast blind restoration algorithm for visual defocused light images based on variational Bayesian is proposed. Through gradient and convolution processing, the posterior probability expectation of the visual defocused light image is calculated, and the optimal initial image and the prior probability of the defocused light blur function are extracted by using the Sobolev space function distribution method. The actual posterior probability is reached infinitely, using relative entropy to calculate the distance between multiple distributions, to approximate the true value of the greatest extent, and input the minimum loss cost function into the bilateral filter, that is, take the approximate clear image as the guide map, to remove the remaining high-frequency noise. The optimal image blind restoration results are obtained. The experimental results show that the proposed algorithm has high image contrast, clear edge details and fast restoration speed after blind restoration, which has extremely high application value.

Aiming at the problems of poor edge security and limited storage effect of unstructured big data, a dynamic secure storage algorithm of unstructured big data based on edge computing is proposed. The unstructured big data is effectively analyzed and identified. The constructed data sensitivity level recognition model is used to establish and encrypt the sensitivity of unstructured big data. Based on edge computing and cloud computing, a cloud edge collaboration architecture is established, and the DCS-SOMP ( Distributed Compressed Sensing Simultaneous Orthogonal Matching Pursuit) algorithm written by the architecture is used to compress and collect encrypted data, so as to reduce data storage. Finally, the unstructured encrypted data is uploaded to each edge of the cloud side collaboration framework to realize the dynamic and secure storage of unstructured big data. Through experimental comparison, it is found that the robustness of storage test, metadata proportion test, encryption time-consuming test and bandwidth consumption is high, which ensures the practical application.

Aiming at the problem that the tracking ability decreases due to occlusion during the tracking of video human moving objects, an adaptive tracking algorithm for video human moving objects considering the occlusion factor is proposed. The Kalman filter and the Meanshift tracking algorithm are used to track the video human moving target. When the target scale, rotation and light change there poor tracking results. The SIFT( Scale Invariant Feature Transform) algorithm is introduced to improve the tracking ability and achieve anti-occlusion video human moving target adaptive tracking. The experimental results show that the tracking accuracy of this method is high, and the tracking success rate can reach about 75% when the occlusion degree is 0. 5. The average tracking frame rate is 28. 9 frame / s, and the real-time performance is strong.