Abstract：DC micro current measurement technology is widely used in nuclear radiation detection, electrostatic detection, space radiation detection and other fields. It is a characteristic extreme value measurement technology, involving many technologies such as high impedance operational amplifier circuit design, leakage current protection and so on. The research on DC micro current measurement technology mainly starts with resistive and capacitive feedback circuits, analyzes the working principle and transfer function of the measurement circuit, focuses on the influence of leakage current, thermal potential, operational amplifier and resistance parameters on the measurement error, deduces the circuit transfer function and simulation analysis, and finally puts forward a scheme to reduce the influence of various parameters on the measurement results. This paper summarizes the DC micro current measurement methods at home and abroad in recent years, and analyzes the current voltage conversion method, current frequency conversion method and modulation, demodulation and amplification method. The current situation of DC micro current measuring instruments at home and abroad is investigated. There are many high-precision weak current measuring instruments in the world, which greatly improves the weak signal measurement.

Abstract：In combination with the standardization needs of the development of the Internet of Things, this paper outlines the international and domestic Internet of Things security standardization organizations. From the perspective of technology and application, the Internet of Things security standards are classified. Finally, this paper designs the framework of the Internet of Things security standard system, and puts forward recommendations for the advancement of security standardization of the Internet of Things.

Abstract：In the past few years, video action classification has gradually changed from manual feature selection to deep learning end-to-end model. This article discusses the traditional action classification method of manually selecting features and the action classification method based on deep learning, focusing on different deep learning methods including convolutional neural networks, recurrent neural network, dual-stream network, long and short-term memory network, etc., and it summarizes the commonly used video action classification data sets, summarizes and prospects the development of video action classification methods.

Abstract：Object detection methods have great value in the application field of video surveillance. At present, it is difficult to realize real-time object detection in resource constrained video surveillance system. A object detection method based on improved YOLOv3-tiny is proposed. Based on the YOLOv3-tiny architecture, the algorithm optimizes the backbone network by adding feature reuse, and a fully-connected attention mix module is proposed to enable the network to learn more abundant spatial information, which is more suitable for object detection under resource constraints. The experimental data shows that compared with YOLOv3-tiny, the algorithm reduces the model volume by 39.2%, the amount of parameters by 39.8%, and improves the mAP of 2.7% on the VOC data set, which significantly reduces the occupation of model resources while improving the detection accuracy.

Abstract：Clinicians need to judge the classification of cystic liver echinococcosis through personal experience when diagnosing liver echinococcosis. The automatic detection and classification model of liver hydatid lesions based on target detection algorithm was studied to realize the automatic recognition and classification of ultrasonic images of liver hydatid disease. The YOLOv5l model was used as the target detection model for cystic liver echinococcosis, and the network model was trained using the local liver echinococcosis ultrasound image data set. The automatic detection and classification model of liver echinococcosis lesions based on YOLOv5L model and stochastic gradient descent algorithm(SGD) optimization algorithm can effectively detect five types of lesions. The average precision mean(mAP) is 88.1%. After testing, the test speed of the model can reach 40 f/s. The experimental results show that the automatic detection and classification model of liver echinococcosis lesions based on YOLOv5L and SGD algorithm can better identify the specific location of lesions and assist doctors in the diagnosis of liver echinococcosis.

Abstract：Vehicle attribute detection is a basic task, which can be applied to many downstream traffic vision tasks. This paper presents an improved vehicle attribute detection algorithm based on YOLOv5. Aiming at the problem of small target detection, this paper adds the convolution attention module to make the network model pay more attention to the small target object. Aiming at the problem of less sample types of the dataset, this paper improves the mosaic data enhancement method of YOLOv5. The self-gated activation function Swish is used to suppress noise, accelerate convergence speed, and improve the robustness of the model. In addition, this paper also makes a detailed vehicle attribute labeling based on the public vehicle dataset VeRi-776, and constructs a vehicle attribute dataset. The experimental results show that the average accuracy of the improved algorithm is 4.6 % higher than that of the original YOLOv5, which can accurately detect the general attributes of vehicle images and can be used for downstream tasks.

Abstract：System level packaging SiP has become the main technical route to break Moore′s law. It is an important support for the miniaturization and multifunction of weapons and equipment in the future. It has a broad market and prospect in the field of weapons and equipment development and application. According to the requirements of lightweight, miniaturization, integration and low power consumption of regional signal target recognition system, a sub component based on SiP technology is designed, and an automatic test system is developed. Through the test of more than 200 products by the automatic test system, it is concluded that the performance of the product is stable and reliable; lightweight, miniaturization, low power consumption and other requirements also meet customer requirements.

Abstract：A high-speed linear push-pull Mach-Zehnder Modulators(MZM) driver based on 0.13 μm SiGe BiCMOS process is proposed in this paper. The driver circuit is composed of 3 components, including the input stage, the output stage and direct current offset compensation(DCOC). A variable gain is achieved by placing variable resistances and variable capacitances in the emitter of the input stage, and the bandwidth is expanded by shunt inductive peaking technology. The output stage utilizes the breakdown voltage multiplier to attain a large-swing, and shunt inductive peaking technology is implemented to get a wide bandwidth. The DCOC builds a feedback between the output and the input of the driver, and utilizes a low pass filter to ensure loop stability. The simulation results demonstrate that the driver gain can be adjusted in a wide scale, and the phase margin, the operating rate and the voltage output swing are 82°, 100 Gb/s, 4 V, respectively, which can drive the push-pull MZM well.

Abstract：This paper presents a design and implementation of a multi 10-bit successive approximation register anglog-to-digital converter(ADC) with a 0.13 μm technology. A C-R DAC optimization technique is utilized in proposed design to reduce the area of ADC. Using thermometer codes to reduces harmonic distortions due to clock feedthrough. Using bootstrapped switch to improve sampling accuracy. The comparator is designed using static architecture with preamplifiers to eliminate offset. The chip is 580 μm×290 μm. The post simulation show that with an sampling ratio of 1 MS/s,200 kHz signal, the ENOB is 9.5 bit and EO=1 LSB.

Abstract：The application of deep learning on electronic design automation of integrated circuits is an interesting topic. This paper investigates the possibility of using Yolo v3 neural network to perform layout checks, which uses the arrangement rules between power and ground pads as inspection cases. In order to generate a training picture set, we use a custom Python script to generate layout sample pictures in batches and utilize LabelImg to label. The Yolo v3 layout checker is written under the TensorFlow framework. Evaluations demonstrate that the proposed layout checker achieves both high accuracy and high recall rate when judging the correctness of the pad layout. Additionally, the inspector is further tested by adjusting the size, shape, symmetry, and number of pads of the layout. Under such circumstances, the inspector still possesses an outstanding performance, showing great scalability. Our research reveals that the Yolo v3 neural network is able to find out errors in pad layout efficiently. Deep learning has great potential in integrated circuit layout inspection, which is worthy of further exploration.

Abstract：For medium voltage distribution network segment positioning method, aiming at the inaccurate positioning problem caused by environmental factors such as the system neutral point grounding way, the size of the distance and the transition resistance, as well as human factors such as current transformer polarity unknown or incorrect erection smart meters and so on, this paper puts forward a kind of stationary wavelet polarity check the fault section locating method based on siamese convolutional neural network(S-CNN). Firstly, the transient characteristics of zero-sequence current are analyzed, and the localization defects of traditional linear correlation method are pointed out. Secondly, the stationary wavelet transform(SWT) is used to solve the problems of signal synchronization and equipment reverse connection. Finally, S-CNN is introduced to perform similarity matching for upstream and downstream signals of the fault point, and the model can be trained to locate the fault segment accurately. The simulation results show that this method has strong anti-interference ability and high recognition rate for blind area.

Abstract：In the research of near-field electromagnetic radiation test, there is no complete set of test methods for individual components. Aiming at this problem, a near-field electromagnetic radiation test method for System in Package(SiP) devices is researched based on surface scanning method. Firstly, the internal structure of the SiP device is researched by X-ray and the source of interference is analyzed. Secondly, the hardware and software layer is built to make the device work. Thirdly, a near-field testing system is built to implement near-field test of DUT. In the case research, the SiP device used is encapsulated with peripherals and a processor which is the main interference source. The near-field test results show that the radiation on PCB is mainly concentrated around SiP device, and the near-field radiation of device is concentrated at processor chip. Case research results show that this kind of test method can effectively measure the SiP device of near field electromagnetic radiation, and analyze the interference sources inside the device.

Abstract：According to the design requirements of satellite X-band communication system link, studying detailly various losses, interferences in the transmission process, system indexes and characteristics，a variety of link calculation methods for uplink/downlink, single/dual channels of satellite communication system are proposed, and the feasibility of the calculation model is verified by practical examples. The calculation method model can be used for parameter selection of satellite communication system, prediction of signal quality and link capability, and it has guiding significance for the design of satellite communication system.

Abstract：In the vehicular ad-hoc network, attackers can publish false traffic information by forging or tampering with messages, etc., resulting in traffic congestion or even serious traffic accidents. However, traditional intrusion detection methods cannot meet the application requirements of vehicular ad-hoc network. In order to solve the problems such as low performance, instability and high storage and time cost of intrusion detection methods in the current vehicular ad-hoc network, this paper proposes an intrusion detection method L-DenseNet(Light Dense Neural Network) based on dense neural network. The L-DenseNet is proposed to reduce the complexity of the model and improve the training speed and deployment adaptability of the detection algorithm. The proposed method is more suitable for intrusion detection in vehicle ad hoc networks. This paper conducts comparative experiments on the VeReMi dataset. The results show that the method proposed has the best overall performance in identifying various types of attacks in terms of precision and recall. As the same time, this method has less time cost and storage overhead.

Abstract：Aiming at the problem that some algorithms are not efficient in the detection of small-scale mobile phone screen defect data, this paper proposes a visual detection algorithm for mobile phone appearance defect based on YOLO v3，namely YOLO-q-v3 algorithm. The YOLO-q-v3 algorithm improves the DarkNet-53 network structure, reduces the number of layers of the original network and the corresponding model parameters, and effectively improves the detection efficiency of the algorithm. The YOLO v3 algorithm and the YOLO-q-v3 algorithm proposed in this paper are compared and tested on the same data set. The experimental results show that the YOLO-q-v3 algorithm can correctly identify the cracks on the mobile phone screen, and the detection rate of YOLO-q-v3 is 24% higher than that of the YOLO v3 algorithm.

Abstract：The city′s 3D model is a bridge connecting physical space and digital space as a three-dimensional visual representation of urban physical entities, and it′s also a key foundation for the realization of urban digital twins. This paper studies the current digital 3D reconstruction technology and application of cities, and summarizes the multi-scale full-element modeling method of "sky, ground and room element". The challenges and trends of digital twin urban 3D reconstruction are analyzed.

Abstract：Post-processing analysis is an important part of computational fluid dynamics simulation analysis. In this paper, a post-processing service model is designed for the cloud computing environment. Based on the model, a post-processing service prototype software is implemented. The user can realize the post-processing analysis of the flow field data through client access. This post-processing service for the cloud computing environment makes full use of the high-performance computing resources of the cloud server to realize the high-efficiency post-processing calculation of large-scale data, realizes the "cloud + terminal" flexible post-processing analysis, and has good scalability.

Abstract：The defect detection of mobile phone cover glass is mainly divided into several steps: image acquisition, image preprocessing, image segmentation and defect classification. Due to the difficulty of obtaining high-quality images, the subsequent image preprocessing will be particularly important. Traditional filtering methods, when dealing with image noise, more or less blur the picture and lose part of the effective information. Usually the presence of noise will make the extreme value in the nearby neighborhood have a large gap between the upper and lower extremes, so filtering becomes an indispensable step. This paper improves the traditional median filter，by judging whether the target pixel needs to be filtered, while enhancing the image, it reduces the loss of useful defect information. This article uses histogram equalization to further process the image after the filtering process, which has the effect of image enhancement. Compared with the traditional median filtering, this method not only preserves the edge information of the flaws, but also has better image enhancement effects.

Abstract：At present, the AIDC interconnection is based on the X.25 line between the civil aviation control units，and the physical line needs to be shared between main and standby automation systems. When the automation system state is switched, it is often necessary to manually switch cable connections.This increases switching time and increases the risk of failure. In this paper, a serial switcher design is proposed, so that the X.25 processing equipment of 2 sets of automation system can share a physical link by switching, and supports three switching modes. At present, the switcher prototype has been tested on the Beijing ACC test platform, and the rapid switching of physical links can be realized in the scenario of AIDC link switching of the automation systems.

Abstract：In order to reduce the phase noise and complexity of the frequency synthesizer, a new design method of low phase noise frequency synthesizer is proposed in this paper. It uses the harmonic generator to generate the high-frequency signal with low phase noise. At the same time, the frequency synthesizer chip LMX2820 integrating the voltage controlled oscillator is used to directly generate the output signal and feedback signal. The feedback signal and low-phase noise high-frequency mixing generate the low-frequency feedback signal, Low phase noise is achieved by reducing the frequency division value through this built-in mixer. The phase noise index of the frequency source of 13.75 GHz~16.25 GHz(excluding 15 GHz) is better than -102 dBc/Hz@1kHz.

Abstract：According to the current requirements for high isolation between frequency bands caused by congested spectrum resources, a new type of miniaturized high suppression waveguide filter is proposed in this paper. Based on traditional waveguide filters, this paper introduces a new CT-type inductive cross-coupling structure between non-adjacent resonators which results in limited transmission zeros in the stopband, and further improves out-of-band rejection, reduces the size of the filter, and reduces the loss introduced by the number of cavities. At the same time, because of the discontinuity of its structure, the mode-matching technique is used to design and optimize the filter. With this method, the design cycle of filter is very short and performance is particularly good.

Abstract：In the background of the operating frequency of 2.4 GHz, based on the designed Doherty power amplifier, an improved Doherty power amplifier with multi-mode switch control and envelope tracking modulation is designed. The tube used in this design is a GaAs transistor NE6510179A from Renesas. The power added efficiency of the designed Doherty peak output(32.0 dBm) reaches 27.2%, and the power added efficiency of the back-off 6 dB is 27.0%. After being improved in cascade with the built envelope tracking model, the new structure can improve the power added efficiency of low output power to a certain extent, and the linearity in the middle and high output power part has been greatly improved.

Abstract：The test of the superconducting quantum computing chip requires dozens or even hundreds of high-frequency microwave signal inputs. The ordinary microwave signal source generally only has 1~2 channels, which is far from meeting the test requirements. If multiple signal sources are integrated to expand the signal channel, it will be not only bulky and expensive, but the synchronization control of the signal source will be also a difficult problem. Aiming at the above problems, based on the power amplifier and power division technology, in this paper，a method of equal-amplitude expansion of one output of the microwave signal source has been proposed, and a signal source equal-amplitude expansion device integrating attenuator, power amplifier and power division module has been designed. Then an experimental platform has been built with microwave signal source, spectrum analyzer, network analyzer and other measuring instruments to carry out experimental research on the performance of the designed signal source expansion device and verify the microwave signal source expansion method based on power amplifier and power division technology. The experimental results show that the proposed method can ensure that the amplitude and frequency of the extended signal are highly consistent with the source signal, and the signal phase of each extended channel can be kept stable for a long time, which meets the requirements of superconducting quantum computing chip testing.

Abstract：The electrical capacitive power transfer(ECPT) system of a single-switch DC-DC converter can increase the operating frequency of the system, but the switching tube is subjected to a high cut-off voltage. This paper proposes a single-switch Buck-Boost radio energy transmission system using LC compensation network. The introduced LC compensation network can reduce the voltage stress of the switching tube. On the other hand, the output is somewhat affected by load changes. At the same time, the converter works in the current source mode, and the coupling power can be used to transmit power efficiently. And the circuit can realize soft switching, and the control of the circuit is easier and safer. Simulation and experimental results show the validity of theoretical analysis.

Abstract：In order to obtain high precision amplitude information of large dynamic range signals to be sampled, a 1 GS/s, 16 bit high-speed and high-precision data acquisition system is designed based on the time-interleaved sampling technology. Through tests, it is found that if the input analog signal contains a large DC component, the output sampled data waveform will oscillate. It is confirmed that the bias error is the cause of oscillation. By setting the ADC to work in the real-time error correction mode, combined with the off-line correction algorithm, the oscillation problem of sampled data is solved, and the time-domain waveform and spectrum before and after correction are compared to verify the effectiveness of the correction algorithms. The test results show that the designed data acquisition system realizes the function of single channel high-precision measurement of large dynamic range signals.

Abstract：In order to realize the rapid and high-precision recognition of air-based downward-looking infrared targets, a single-stage space-based down-looking multi-angle infrared target recognition algorithm is proposed. Firstly, use Darknet-53 combined with SPP module to perform feature extraction on infrared targets, to fuse local features and global features to improve the expression ability of feature maps, and finally use Focal loss in RetinaNet to lock the detection box of the target, and at the same time obtain the target type and detection accuracy. Aiming at the defect that the existing data sets are mostly head-ups and single viewing angles, a composite-wing drone was used to collect infrared images from different heights and angles, and a multi-scale down-view infrared target data set was constructed, which was implemented and verified on the PyTorch architecture. The proposed algorithm achieves 91.74% of the mAP of the downward-looking infrared target, the recognition speed is 33 FPS, and it also has a good recognition result on the public infrared ship data set. The experiments show that the algorithm meets the real-time requirements on the basis of ensuring accuracy, and provides theoretical technology for subsequent real-time recognition of multi-scale targets on compound-wing UAVs.

Abstract：In recent years, China′s independent software and hardware joint ecology has developed rapidly. The establishment of the hardware and software ecosystem includes the infrastructure composed of the underlying hardware and operating system, abundant general tool software, a large number of middleware, applications, and a large number of developers. In order to solve the healthy development of independent hardware and software ecology, it is necessary to understand the development of ecology and study its construction mode. This paper innovatively summarizes the current construction of independent software and hardware ecology in China into three modes, namely product mode, group mode and service mode. By combining the adaptive product data and developer behavior data, the characteristics and differences of the three construction modes are expounded, providing experience reference and decision support for the successful construction of the independent software and hardware ecosystem in China′s IT industry.