Abstract： With the development of agricultural big data and smart agriculture, in the face of massive agricultural text data, the demand for building knowledge graph and other natural language processing applications has gradually increased. At present, the entity corpus and entity labeling system in the agricultural field are still in a blank state. When dealing with agricultural texts, we are faced with such problems as how to define the category and scope of entities. Based on this problem, this paper takes the agricultural thesaurus as the scientific basis, proposes the agricultural text data entity labeling criteria for the construction of agricultural knowledge graph, covering a variety of agricultural entities such as crops, pests and weeds, and constructs a self-annotation corpus based on agricultural text based on the labeling principles of the criteria, and carries out experimental verification to prove the effectiveness of the criteria. This criterion provides a referential labeling specification for the construction of agricultural entity corpus and corpus support for agricultural entity recognition.

Abstract： Fruit fly is a kind of quarantine pest that attracts much attention at home and abroad. There are many kinds of fruit flies. Different kinds of fruit flies are similar in shape and size, which is difficult to identify. In addition, in practical applications, it is difficult to identify fruit flies due to the lack of information about shielding, view-point, changing light and shadow and other factors. This study proposes a bilinear pooled attention network for fruit fly classification to learn effective discriminant characteristics. The network is composed of two parts: saliency feature module and cross-layer bilinear feature module. Saliency feature module realizes feature enhancement by filtering enhancement processing of two different convolution layers. Cross-layer bilinear module is based on bilinear pooling fusion features, determines the attention location, and mines discriminant features. Experiments on fruit fly’s data set with natural environment background show that the method is effective and has good practical application prospect.

Abstract： Chlorophyll fluorescence imaging, as an efficient means of obtaining plant photosynthesis and physiological state information, is often used in many fields such as smart agricultural information perception. The current large-scale chlorophyll fluorescence imaging technology lacks the mechanistic connection with the microscopic leaf-level chlorophyll fluorescence imaging, ignoring the impact of imaging angle and leaf age on chlorophyll fluorescence imaging, which limits its application in large-scale agricultural intelligence assessment and role in decision-making. Aiming at the above problems, this paper designed a set of chlorophyll fluorescence acquisition equipment capable of multi-angle imaging to explore the relationship between chlorophyll fluorescence imaging, imaging angle and leaf age.A total of 60 leaves of epipremnum aureum,camphor and euonymus, representing vines, trees and shrubs were subjected to imaging experiments, and the fluorescence intensity, F_v/F_mand R_fd three typical fluorescence indicators were analyzed. The experimental results show that the imaging angle has a significant impact on the chlorophyll fluorescence imaging, and the increase of the imaging angle will lead to the decrease of the fluorescence parameter value, and the effect on different types of plants is also different. Leaf age also affects chlorophyll fluorescence imaging, and the imaging parameters of growing leaves are better than those of mature leaves.

Abstract：Global digitization has entered a period of accelerated development, and China's digitization process is in full swing. The deep application and wide penetration of the new generation of information network technologies represented by the Internet of Things, big data, cloud computing and artificial intelligence in the agricultural field has a profound impact on the innovation and upgrading of the agricultural industry. Therefore, digitalization and intelligence of agriculture is an inevitable trend of modern agriculture development, and an inevitable measure for Chinese agriculture to plug in the wings of science and technology. Planting industry is related to the national economy and people's livelihood. The digitalization and intelligence of planting industry is the central link of the digitalization construction of the whole agricultural industry chain and the main component of modern farm.

Abstract：Facing wide-area and non-uniformly distributed service requirements, this paper proposes a beam resource management scheme based on improved particle swarm optimization, which considers the joint optimization of beam hopping pattern design, beam frequency allocation and power allocation, in the scenario of multiple LEO satellites with moving beam cells. By system-level simulation, its performance is verified. Specifically, compared with beam resource management schemes based on spatial isolation angle and priority, results show that this proposal can significantly improve user SINR and system capacity under full-buffer service model, and lower frequency reuse factor results in higher system capacity.

Abstract： As an important feature and key enabling technology of 6G, integrated sensing and communication (ISAC) achieves information sharing and collaborative gain for communication and awareness. This paper firstly introduces the current research status of ISAC technology, and then combs it from two aspects: academic research and standardization progress. Then, this paper deeply discusses and summarizes the integration design issues of communication technology and sensing technology in three aspects: network architecture, networking and waveform design. Finally, the development trend of ISAC technology in 6G is forecasted and the possible challenges are analyzed.

Abstract： NTN (non-terrestrial network) is an important application scenario for satellite communications and low-altitude communications, marking the transition of 5G technology applications from land communications to space communications. It is foreseeable that satellite networks will be an important component of future 6G communications networks. In order to meet the quality requirements of satellite communication and maximize the system capacity, it is necessary to apply adaptive coding and modulation technology to dynamically adjust the modulation order and coding rate according to the channel state information in the changing communication environment. AI has clear potential to solve problems arising from rapidly changing channel conditions in satellite high-dynamic scenarios. This paper adopts the low-orbit satellite adaptive coding and modulation strategy based on reinforcement learning to solve the problem of the mismatch between the threshold table and the actual channel caused by the change of the satellite communication environment, which is improved by above 20% compared with the traditional ARIMA (autoregressive integrated moving average) algorithm.

Abstract：Autoencoder (AE) based on deep learning is a new method to replace traditional communication transmitter and receiver. This paper proposes an autoencoder based on Convolutional Neural Network (CNN) and Gate Recurrent Unit (GRU), which integrates constellation mapping and channel coding. Specifically, this paper designs a parallel CNN structure and segment the input bitstream for one-hot encoding, which has two advantages:(1) Compared with the original one-hot encoding, the dimension of the input data is reduced; (2) The features of the data are not too sparse, which allows the network to converge faster and better. In addition, the GRU is introduced for channel coding. The proposed model can be applied to high-order modulation such as 4096QAM signal, and has better performance than traditional methods under both added white Gaussian noise (AWGN) channels and Rayleigh channels.

Abstract：Data trading plays a vital role in the development of digital economy. This paper centers on the overall requirements of data element market establishment, as well as the practical basis and overall requirements of a national unified market. It sorts out the pain points exposed in current data trading, and puts forward the idea that in order to solve the dilemma of data trading and accelerate building a national unified data trading market. It is essential to solve four pairs of relationship properly: security and development, government and market, competition and cooperation, and management and technology. It then points out that to promote the establishment of a unified data trading market, four issues need to be focused on: establishing a unified trading system, promoting the complementary functions of trading venues, strengthening infrastructure connectivity and promoting business separation and ecological establishment.

Abstract：The Scene Graph Generation (SGG) task aims to detect visual relation triples in images, i.e. subject, predicate and object, to provide a structural visual layout for scene understanding. However, existing approaches to scene graph generation ignore the high frequency but uninformative problem of predicted predicates, hindering progress in this field. In order to solve the above problems, this paper proposes a scene graph generation algorithm based on enhanced semantic information understanding. The whole model consists of four parts: feature extraction module, image cropping module, semantic transformation module and extended information predicate module. Feature extraction module and image cropping module are responsible for extracting visual features and making them global and diverse. The semantic transformation module is responsible for restoring the semantic relationship between predicates from common predictions to informative predictions. The extended information predicate module is responsible for extending the sampling space of the information predicate. Comparing with other methods on datasets VG and VG-MSDN, the average recall reaches 59.5% and 40.9%, respectively. The algorithm in this paper can improve the problem of insufficient information of the predicted predicate, and then improve the performance of the scene graph generation algorithm.

Abstract： Feature extraction and analysis of medical text is of great practical value in building clinical decision support systems. A medical text analysis model based on BERT and Word2vec is proposed for the situation that raw medical texts containing various terms and abbreviations are difficult to extract features. The model extracts key medical entities from medical records and establishes a weight scoring mechanism based on knowledge for semantic analysis of medical texts. The experimental data show that the model has certain advantages in medical text feature extraction, good performance in the analysis and diagnosis of hypertensive intracerebral hemorrhage medical records, and can be effectively used in clinical decision support systems.

Abstract：In the vector control of the permanent magnet synchronous motor, it is necessary to obtain the accurate electrical angle of the motor rotor. In industrial control, the motor position sensor often uses a resolver to obtain the motor rotor position. Decoding the resolver position usually uses a dedicated decoding chip to obtain the rotor position of the motor, but the decoding chip increases the hardware development cost. In this paper, the software algorithm design and analysis for the position decoding of the resolver is carried out. The traditional second-order angle tracker has a poor acceleration tracking effect and is optimized, and a feedforward correction device design for the second-order tracker is proposed to achieve better angle tracking effect. The software algorithm is simulated and verified by MATLAB/Simulink. The MCAL is used to configure the TC275 peripheral DSADC to decode the resolver information, and the feasibility of the scheme is verified through experiments.

Abstract： For the actual power supply circuit system, the current detection circuit is designed to realize the real-time detection of each power supply current. In the process of current detection circuit design, it is usually necessary to select and calculate the resistance values of the sampling resistance and conditioning resistance. In view of this complicated and highly repetitive work, the technology of using Excel VBA to compile resistance selection calculation module including series, parallel and proportion is proposed. It can quickly and effectively solve the problem of resistance selection in current detection circuit. The results show that using VBA program to calculate the resistance data can automatically generate the target resistance value on the Excel worksheet cells. This method can effectively improve the accuracy and efficiency of the resistance selection calculation of the sampling resistance and conditioning resistance.

Abstract：With the increase of single high-rise city buildings, fire rescue has extended from simple plane rescue to three-dimensional space. Frequent fires pose a great threat to the life safety of fire rescue personnel entering such fire sites. Locating the location of people on the fire scene for rescue when needed becomes an extremely valuable subject. Based on ultra-wideband (UWB) positioning technology, since the Least Square Error/ Time Differential of arrival (LSE/TDOA) hybrid algorithm only has a good performance in two-dimensional solution, but the three-dimensional positioning accuracy is not high, and the error is large, the Chan- Weighted Least Squares (WLS)/TDOA positioning algorithm is further proposed. The experiment proves that the algorithm has high stability and positioning accuracy. At the same time, the Cramer-Rao lower bound(CRLB) of the 3D TDOA algorithm is theoretically deduced for the first time, and the CRLB value of the proposed Chan-WLS/TDOA positioning algorithm is obtained. Experimental results show that the errors are under control when adapting the positioning algorithm in a variety of complex environments and can meet the needs of engineering applications.

Abstract： In order to solve the problem of continuous error accumulation caused by propagation in the process of anchor-free cooperative localization, an improved map-growing algorithm based on mesh network is proposed. Firstly, the geometric collinearity formula of virtual anchor node is presented by deducing the estimation value of positioning error between reference nodes and node to be tested. Then, the virtual reference nodes are selected according to the multi-condition constraints to establish the relative reference coordinate system, and the optimal node set is used to complete the cooperative localization. Finally, the node uncertainty is introduced to accomplish the node upgrading strategy, which can curb the spread of cumulative error in the iterative positioning process. Simulation results in various node distribution scenarios show that the proposed algorithm effectively alleviates the accumulated errors in the collaborative positioning process and achieves good positioning performance.

Abstract：The history station undertake a task of collecting,storing and querying in the DCS(distributed control system),moreover the real time database and history database are key technology in the configuration software. In this paper, a design way of the real time database and the history database based on the double history station is studied. It introduces the general structure of the distributed control system, the data acquisition，the memory structure design of the real time database, the design of the history database, the synchronization policy, and the extern interface of the database. The database system has been widely applied to the multiple fields of the industrial control, which has the characteristics of high real-time performance, less storage space and fast query.

Abstract： The traditional stepper motor controlled by pulse command has the characteristics of open-loop positioning without any position sensor. The disadvantage of open-loop control is that it is easy to lead to out of step when overload occurs. In order to reduce the probability of out of step, the stepper motor usually works with the maximum current, which will affect its working efficiency. To solve the above problems, a control algorithm based on the factor of (Δcos(ϕ))⁄ΔI to optimize the driving current of the stepper motor has been proposed in this paper. The algorithm combines the characteristics of stepper motor and its pulse command, has strong applicability and simple calculation process, and doesn′t need to set additional mechanical or electrical parameters of the stepper motor. Using the measured data of hybrid stepper motor, data simulation experiments have been carried out to prove the effectiveness of the proposed algorithm.

Abstract：Different from the optical imaging system, the imaging system of optical microscope has a small depth of field. The main core problem in the focusing process is the selection of focusing evaluation function. In this paper, eight commonly used focus evaluation functions are tested, and the focus evaluation curve is qualitatively analyzed. The sensitivity factor, sharpness ratio, fluctuation in gentle area and running time are used for quantitative analysis. Combined with the analysis data, the optimal focus evaluation function suitable for micro imaging system is selected.

Abstract： In order to reduce helicopter rotor noise, the model of helicopter rotor noise active control system is established by using FXLMS algorithm. The off-line identification method is used to identify the secondary sound path. The rotor speed signal is used to construct the sound source reference signal, so as to avoid the interference of the secondary sound source to the reference signal. Combining normalization algorithm with G-SVSLMS algorithm, an improved variable step size algorithm is proposed. The algorithm not only has the advantages of fast convergence and small steady-state error, but also adapts to the time-varying characteristics of the reference signal, and the parameter selection is convenient. The rotor noise is simulated in this paper. Compared with G-SVSLMS algorithm, this algorithm has faster convergence speed, better adaptability to the change of reference signal, and smaller steady-state error compared with normalization algorithm. The results show that the algorithm can effectively reduce the helicopter rotor noise, and the noise reduction effect can reach 24.8 dB at a certain frequency. At the same time, the stability of the system is improved.

Abstract：The communication system based on auto-encoder is a hot research field of communication system in recent years, and how to apply the end-to-end learning communication systems to embedded devices is a hot topic worthy of research. This paper proposes a FPGA design scheme for end-to-end wireless communication system based on auto-encoder, deploys the end-to-end wireless communication system based on auto-encoder on the FPGA, and uses the AD9361 radio frequency chip as the radio frequency front-end processing module to achieve real air transmission. In addition, a hardware acceleration scheme is designed for the CNN in the system, parallelism is explored in the convolution computing unit, and a pipeline architecture is designed to accelerate the convolution operation. For the storage unit, the double buffer design is adopted, and the ping-pong operation is used to improve the utilization rate of data communication bandwidth. The experimental results show that under different modulation methods, the BLER performance is close to the simulation results under the Rayleigh fading channel. Compared with the general-purpose CPU i5-9300, the network inference speed of the designed system is 3.98 times higher than that of the general-purpose CPU i5-9300 with the same block error rate. Compared to the Nvidia 1650 GPU, the power is about 0.18 times higher.

Abstract：Wireless power and data transmission technology can transmit information while energy is transmitted wirelessly, so it has been widely studied in recent years. In this paper, binary frequency shift keying (2FSK) is used to modulate the data to the energy transmission channel. The modulation of binary information can be achieved by changing the working frequency of the inverter. The frequency tracking characteristics of the phase-locked loop are used in data demodulation, that is, when the phase is locked, it can output effective signals, and when it cannot be locked, it can output signals with opposite levels. The effectiveness of the wireless power and data transmission technology is verified by the prototype. The wireless power transmission method proposed in this paper can achieve the communication rate with 10 kb/s while the power transmission is normal, and the communication performance will not be affected when the power transmission is changed from tens of watts to hundreds of watts.

Abstract： An “EIE” magnetic integrated transformer made up of different soft magnetic materials is proposed for LLC resonant converter with power factor correction function, which integrates PFC inductor, LLC resonant inductor and LLC high-frequency transformer by decoupling integration. The design method of magnetic integrated transformer is introduced in detail, and the transformer is modeled and simulated by ANSYS Maxwell simulation software. For improving the anti-saturation ability of transformer, I-type magnetic core material is selected as Fe-Si-Al magnetic powder core after comparing the characteristics of Mn-Zn ferrite material and Fe-Si-Al magnetic powder core material. Finally, the effectiveness of the design is verified through an experimental prototype with 100 W power.

Abstract： A loop compensation circuit for voltage control mode of buck-boost switching power supply is designed. This compensation circuit is TYPEIII compensation loop. This paper analyzes the power level modeling of buck-boost switching power supply and the circuit design of error amplifier compensation loop to ensure the stability of switching power supply. The proposed circuit is based on 0.5 μm process for layout design and simulation verification. The results show that the bandwidth and phase margin of the error amplifier meet the requirements. The voltage regulation of the system is 0.42%, the stability time is 300 μs, the load regulation is 0.17%, and the stability time is 300 μs. It contents the design requirements and applies in a buck-boost switching power supply chip.

Abstract：The flyback converter topology is mostly used for low-power power supply, and in the case of high-voltage input, switching voltage spikes of power devices have become more serious, and even cause the chip to fail to work normally. In this paper, the generation mechanism of flyback converter noise is analyzed in detail. Combined with transformer leakage inductance regulation and converter load conditions, the output voltage spikes can be suppressed. Secondly, using the characteristics of TL431 and optocoupler converter, a variable compensation network is designed to compensate system performance degradation caused by leakage inductance change from loop control. Finally, a multi-output flyback power with input 430~650 V, output 10 W is built. The experimental results verify the feasibility and correctness of the design.

Abstract： A local positioning system based on ultra-wide band (UWB) technology is proposed to realize unmanned aerial vehicle (UAV) navigation in the environment of satellite denial. Firstly, the UWB technology is used to realize the ranging, and then the three-dimensional coordinates of the UAV in the local coordinate system are solved. Then the multi-sensor data are fused based on extended Kalman filter (EKF) to improve the positioning accuracy of the UAV. Based on the open source flight control Pixhawk, the UAV flight verification platform is built, and the UWB positioning system data is fused with the flight control sensor data to simulate the autonomous flight of UAV in the local coordinate system. Finally, through experiments, the feasibility of multi-sensor data fusion navigation based on UWB in the autonomous navigation control of UAV is verified, and the flight navigation tasks of UAV such as autonomous take-off and landing on the ground platform and flying with ground targets are realized.

Abstract：Under the circumstance of GPS-denied environment, a visual positioning method based on high-definition aerial images is proposed. In order to ensure the real-time performance of high-resolution image matching, two-stage fast scene matching is realized by combining coarse-grained high-level semantic global features and fine-grained point features. In the first stage, a triplet network is designed to extract scale and rotation-invariant global descriptors for referance image blocks for region pre-matching. In the second stage, Ann-SIFT for fast nearest neighbor search performs precise point matching on the real-time image and reference area to achieve localization. For reference images of 3 000×3 000 and real-time images in the size range from 500×500 to 1 000×1 000, the average time-consuming can be reduced from 66.24 s to 0.97 s.

Abstract：With the improvement of the resolution of synthetic aperture radar (SAR), the use of SAR images for ship detection and identification has gradually become an important means of marine target surveillance. However, limited by the SAR imaging mechanism, the side-lobe problem of high-resolution SAR images is becoming prominent gradually, which seriously affects the subject segmentation of ship targets. In this paper, a fine segmentation algorithm of ship target based on Radon transform is proposed. By performing Radon transform on SAR images, the identification and elimination of sidelobe are realized in the Radon domain. Then, the morphological filtering is used to remove the fine side lobes. Finally, the effective suppression of the SAR image side lobes is realized. The algorithm is verified by GF-3 and COSMO-SkyMed satellite image data. The results show that the algorithm has better performance in uniformity of intra region, dissimilarity of inter region, and complexity of shape compared with existing segmentation algorithms.