In recent years, the research of liquid crystal microwave communication devices has developed rapidly. The high dielectric loss of liquid crystal materials has become a bottleneck restricting the development of microwave devices. There are few reports on the properties of liquid crystal materials at microwaves. Therefore, we take the fluorine substituted tolane-type compounds with low melting point and high birefringence as the research object, doping them into the liquid crystal matrix MA in a certain proportion respectively, and testing them at microwave k-band (10~30 GHz) through the rectangular cavity perturbation method. And then, the influence of molecular structure on the dielectric properties of liquid crystals at microwave frequency is discussed.The experimental results show that dielectric anisotropy at high-frequency is related to the polarity of liquid crystal molecules and birefringence. The lateral fluorophenyl phenylacetylene and terminal isothiocyano phenylacetylene liquid crystal compounds exhibit high dielectric anisotropy(Δεr>0.85). Triphenyl diacetylenes substituted by F-atom symmetrically and triphenyl acetylene liquid crystal compounds with terminal isothiocyanate exhibit lower dielectric loss(tanδεr⊥0.008, 18 GHz).

Liquid crystal (LC) is a kind of functional soft matter featured by self-assembly and stimuli responsiveness. It is widely known as a fundamental material for current mainstream information display industry. With the development and innovation of LC theory and technology, understanding on such a material system has been significantly extended. Recently, the manipulation of LC microstructures has been flexibly enabled via a photoalignment technique, and the research focus has been gradually transitioned from traditional displays to advanced LC photonics. It is expected to provide great opportunities in the fields of planar optics, structured light fields, optical cross connection, and mode division multiplexing optical communications. This review presents the latest works of the Research Center for Liquid Crystal and Photonics of Nanjing University in the field of photoalignment-enabled LC microstructures and photonic applications. To be specific, LC hierarchical architectures, optically addressed spatial light modulators, and LC telecom/terahertz elements are separately discussed.

Mini-LED backlight can realize local dynamic dimming through refined partitions, which is a research hotspot of current Liquid Crystal Display(LCD) backlight technology. This paper aims to reveal different impact on the optical performance and system architecture of the backlight module by different Mini-LED local dimming unit types. We establish three different optical models of the Mini-LED local dimming backlight unit, namely the direct-lit dimming unit, edge-lit dimming unit, and the corner-lit dimming unit proposed in this paper. Through optical modeling and simulation tracing for the optical models with Mini-LED light sources in different structures, the uniformity and light efficiency of the dimming unit are compared. The results show that the light uniformity is 6727% and the light utilization rate is 86.67% of the corner-lit dimming unit without optical films. The light uniformity is 84.70% and the light utilization rate is 92.10% of the corner-lit dimming unit with optical films. The optical performance has improvement in different level compared with the edge-lit local unit and direct-lit local unit. This results have important theoretical guiding significance and application value for the design of Mini-LED backlight units.

To improve the reflectivity of vertical alignment Liquid Crystal on Silicon (VA-LCoS), we built a three-dimensional (3D) optical modeling for the normally black VA-LCoS mode, and investigated the effects of the gap of liquid crystal cell, the pretilt angle of liquid crystal and wavelength of incident light on reflection for linearly polarized light and circle polarized light as incident light, respectively. As a result, the black zone in the on-state pixel was eliminated by using circle polarized light as incident light. Moreover, the VA-LCoS has the highest reflection as the cell gap is 1.6 μm. The results have an important guiding significance for the design and preparation of VA-LCoS.

In order to solve the problem that the monochromatic aberration of the electrowetting liquid lens focusing system changes greatly with the focusing process, a design scheme of correcting the monochromatic aberration of the system by using liquid crystal lens with adjustable gradient refractive index distribution is proposed. At 850 nm wavelength, liquid crystal lens is used to correct the monochromatic images of the system at six typical object distances ranging from infinity to 200 mm. At the infinite object distance with the largest aberration, the geometric radius of the dot pattern of the edge field of view is reduced from 44.696 μm to 7.423 μm, and the speckle in other fields is reduced to the Airy spot. And the monochromatic aberration of the system has been corrected. In order to reduce the response time of the liquid crystal lens, the liquid crystal lens is replaced by laminated combination of five liquid crystal lenses with 1 / 5 cell thick, which enhances the response speed significantly. It has been found that this structure also has a favorable correction effect on the monochromatic aberration of the system at different object distances.

When the COB LED display screen is black, the ink color difference between COB LED modules may appear. In order to ensure the consistency of ink color on the same side of COB LED display, this paper studies the ink color collection, classification and sorting algorithm of COB LED display module. Firstly, the ink color acquisition method based on machine vision COB LED display module is introduced, and then, the ink color acquisition device aiming at the optimal effect of viewing COB LED display screen ink color by vision is developed. The COB LED display is completed by using the ink color classification and sorting algorithm based on machine vision. Finally, through the engineering case experiment, the feasibility of COB LED display module ink color collection, classification and sorting scheme based on machine vision is verified, which meets the consistent feeling of human vision as observing a COB LED display.

The direct current regulated power supply adopting the mechanical buttons programming mode has problems, such as easy wear of buttons and complex circuit composition. To solve the above problems, a system is designed with the programmable digital power supply TFT liquid crystal display (LCD) screen touch display based on STM32F7. In terms of the hardware of the system, it is composed of an STM32F7 processor, TFT LCD touch screen, SDRAM chip, and power supply module. STM32F7 is used to drive TFT LCD screen and to expand external SDRAM as video memory through LTDC, and the power supply module provides backlight voltage power for the display screen. The software uses the real-time operating system of FreeRTOS for managing and dispatching system tasks, utilizes emWin for interface design, and adopts DMA pattern to read data. The display system has a simple circuit, stable and reliable display performance, and can control the power system by touching. The experimental results show that the system has a clear display effect, fast control and response. Additionally, the system realizes the human-computer interaction function of the programmable power supply by utilizing the TFT touch screen, thereby providing a reference for the application of the TFT touch screen in the power supply field.

As a primary level of Micro-LED, Mini-LED backlights are emerging as an advanced TV technology of “high-end direct upgrade”. Currently, many consumer electronics companies launched Mini-LED backlight embedded products, including high-end displays, notebooks, tablet computers, car displays, and so on. Motivated by the promising future, there are many suppliers working on the commercialization of Mini-LED technology related components including chips, packages, panels, and complete machines. This review describes the key technology of Mini-LED backlight technology and the status of current commercialization. This paper also discusses the prospect and challenges of TV technology in future developments.

In order to eliminate the image sticking in high-temperature of TFT-LCD, a method to improve image performance by adjusting the positive and negative Gamma voltage and a quantitative research is conducted. The method of asymmetric adjustment of positive and negative Gamma voltage is quantified. Aiming at the problem of image sticking in the LCD field, an electrical principle is analyzed, which is the reason why the image sticking is inevitable in LCD, and a method to quickly adjust the positive and negative Gamma voltages is proposed. The Flicker value is measured in a normal or high temperature environment to determine the variation of Vcom voltage. According to this variation, the positive and negative Gamma voltage of the LCD are ajwsted, so as to realize the quantitative theoretical value to adjust Gamma voltage, and finally achieve the goal of eliminating the residual image. The experimental result shows that the residual image level is reduced from L1 to L0.5 by this method.

In order to eliminate the Mura defect of VA-LCD, the image processing technology is adopted to collect display data through camera, so as to capture display pixel brightness. In-depth research is conducted on the algorithm of camera data correction. Firstly, the fixed mode noise and time-varying noise of the camera are reduced, and the plane field of the camera and lens is corrected by the standard light source. Then, the circular dot matrix pattern is used as the auxiliary calibration diagram, and the correction model is established by using the two-line relation, and the pixel brightness matrix is obtained preliminarily. According to the visual angle characteristics of VA-LCD, the characteristic curve of luminance-visual angle under different grey levels is established, and a large number of data are fitted with a fourth-order polynomial to improve the stability of the visual angle model. Finally, the effectiveness of the algorithm is verified by combining the compensation value calculation and chip real-time compensation technology, and the evaluation is carried out from the subjective and objective indicators. The experimental results show that the JND index of the display decreases from 2.7 to 1.8, and the in-plane 9-point uniformity increases from 53.4% to 82.9% after the application of this Mura elimination technique. The algorithm proposed can meet the requirements of the highest specifications of products and has good robustness in adaptive correction to meet the requirements of mass production in factories.

In recent years, the traditional aerial image target detection algorithms have been unable to meet the requirements of detection accuracy and speed, while the rapid development of target detection algorithms based on deep learning provides a new idea for target detection. However, this kind of algorithm is often accompanied by large scale and highly dependent on GPU devices, which makes the migration of the mobile end of the algorithm difficult. Aiming at the above problems, this paper proposes a MPSOC platform implementation scheme based on Yolo V3 algorithm. Firstly, the anchor frame of the original network is re-selected by means of k-means clustering, the detection accuracy of the algorithm is increased by adjusting the convolutional layer, and then the model scale is compressed by sensity-based pruning operation. Finally, VISDRONE data set is used to verify the Xilinx ZYNQ series MPSOC platform. The experimental results show that MAP of the improved Yolo algorithm increases by 1.3%, and the false detection rate is also greatly reduced. After the model is compressed, the detection speed is doubled and the volume becomes 37% of the original. It basically meets the design requirements of aerial image target detection, and provides a feasible solution for the implementation of deep learning algorithm in MPSOC.

In a complex construction environment, the helmet wearing recognition algorithm based on machine vision technology often fails and misdetects, and its recognition ability is limited. In order to improve the accuracy of helmet wearing recognition, this paper establishes a helmet recognition convolutional neural network based on the bidirectional feature pyramid of the attention mechanism isproposed. In order to improve the expression ability of shallow position information and deep semantic information in the convolutional neural network, and increase the recognition rate of vague and small helmets, the network introduces the jump connection and the attention mechanism CBAM technology into the bidirectional feature fusion feature pyramid network PANet module, and bidirectional feature pyramid module based on the attention mechanism is constructed. In order to improve the convergence ability of the model, CIoU is used instead of IoU to optimize the anchor frame regression prediction, which reduces the complexity of the network training. The results of comparative experiment show that the mAP value of our proposed recognition network is 0.82, 4.43, 23.12 and 23.96 higher than that of YOLOv3, RFBNet, SSD, Faster RCNN, respectively, and its detection speed reaches 21 frame/s, thus satisfy with height real-time accuracy of helmet recognition in the construction environment.

Face detection is the first step of face recognition. In order to fix the face region from the complex background rapidly, a method based on LVQ is studied in this paper. By analyzing the color features of skin, images are converted into YCbCr space and HSI space firstly, and then Cb, Cr, H, S color components are extracted to describe the facial image features. An LVQ neural network with a structure of 4-20-2 is constructed, 100 images as training samples and 20 images as test samples are selected for this neural network. The number of iterations is 150 and the error is 0.001. After training, the effective weight are obtained. Using the trained neural network, the performance tests are performed on the LFW, Faces, and AFW datasets. The positive detection rates in the three data sets are 76.82%, 84.42%, 100%, the false detection rates are 17.34%, 12.34%, and 0, and the missed detection rates are 21.55%, 15.63%, and 0, respectively. The experimental results show that the method in our paper shows good superiority in terms of positive detection rate, false detection rate and missed detection rate compared with the traditional face detection method based on skin color features.

In order to overcome the on-orbit single-event flip and realize flexible and reliable multiple working modes, a refresh imaging system based on Xilinx field programmable logic devices and refresh chip of Fudan Microelectronics group company limited is designed. Using a single set of daisy chain structure serial bus to control the serial port of 5 sets of refresh imaging unit, FPGA internally converts the received data and then configures or refills the refresh chip. The refresh enable signal is used to control whether the refresh chip is powered on, and at the same time decides whether the FPGA is loaded from PROM in the master and parallel mode or loaded under the control of the refresh chip in the parallel mode. Through the control level setting, the selection of the loading data source is realized. After each power supply is completed, the external reset signal is used to refresh the chip reset, meets the timing requirements of refreshing the chip to start configuration immediately after reset. The experimental results show that the transmission time of a single packet of 204 byte data is only 0.377 6 ms with the baud rate of 625 kbit/s in daisy chain mode, which is much lower than the 2.01 ms required to refresh the serial port, and multiple groups of time-sharing work can be performed. The refresh imaging system using a small number of single-ended control signals combined with a daisy chain serial bus has a variety of working modes that are stable and reliable, with design flexibility and on-track maintainability.

The verification of IP core of image processing algorithm is the key of SoC and FPGA application in the field of machine vision. In order to shorten the verification cycle, based on ARM+FPGA heterogeneous platform and combined with software on PC, a verification framework with universal application, real-time performance and agility is designed for IP core of image processing algorithm. The verification framework establishes gigabit ethernet communication with PC through ARM to realize real-time input of test image and video, uses FPGA to build data bus and configuration bus for the purpose of compatibility with multi-type images, and combines partial reconfiguration to realize fast iteration of IP core of the algorithm to be verified. The validation framework is reusable for the algorithm IP core with 8, 16, and 24 bit depth images as processing objects and results, and the deployment speed of the IP cores to be verified is 25 times faster than global reconfiguration. Compared with the existing FPGA verification technology, it has better reusability, shorter verification cycle, more targeted test stimulus, and faster deployment of IP cores to be verified

A novel low-dose reconstruction algorithm using dynamic optimization design was developed to suppress artifacts and noise in computed tomography (CT) imaging under the low-dose scan protocols. In this paper, a family of hyperbolic tangent functions was selected to build the composite function model of fractional total variation (TpV). Furthermore, a dynamic optimization (DO) term was also applied to improve the performance of the presented model. In our method, the proposed iterative reconstruction was achieved via the statistical iterative reconstruction (SIR) framework. Then, the presented approach was evaluated by using X-ray low dose projections collected from simulated phantom and scanned mice. The simulated results for 180 sampling views show that the signal to noise ratios (SNR) of images reconstructed by the proposed algorithm are 29.51, 8.03, 9.15, 6.81 dB higher than those of images reconstructed by filtered back-projection algorithm, TpV algorithm, and adaptively reweighted total variation algorithm, respectively. And the mouse data studies demonstrate that the proposed method suppresses artifacts or noise successfully and preserves more soft tissue details in reconstructed images. Moreover, the quality of reconstructed images can be greatly improved by the presented low-dose reconstruction algorithm.