Rapid development of flexible electronics requires thin film transistor (TFT) to have features of low power consumption, good bending resistance and ability to be fabricated at low temperature. Gate insulator is one of the key materials in TFT device. Organic gate insulator prepared by solution method has the advantages of low cost, strong flexibility and is suitable for large area processing. In this paper, polyvinylpyrrolidone (PVP) dielectric film was achieved by inkjet printing process. The composition difference of printed PVP films with different annealing temperature was analyzed by XPS, and leakage current, capacitance and transfer characteristic of PVP devices were obtained by semiconductor analyzer. TFT device with PVP film annealed at 200 ℃ showed leakage current density ≤10-4A/cm2 (5 V) and relative dielectric constant about 3.8. Saturation mobility of device based on glass substrate is 4.6 cm2/(V?s) and the Ion/Ioffratio is ≥105 while that for device based on PI substrate with bending radius of 20 mm is 2.8 cm2/(V?s) and 6×104, respectively. This result demonstrates that PVP dielectric is promising in flexible electronics.

Electrochromic (EC) devices possess the advantages of low operating voltage, many stable states, non power consumption under static state, low power consumption for color change and continuously adjustable transmittance/reflectance and so on. WO3 electrochromic devices have problems such as long response time and low optical modulation amplitude. In order to further shorten the response time, this article uses WO3 film as the color changing layer to prepare electrochromic devices. The effect of different argon-oxygen flow ratios on the WO3 film deposition and the response time of EC devices are studied. In order to further shorten response time, we used the Ag-WO3 composite layer as the color-changing layer for the EC devices. The results show, when the argon-oxygen flow ratio is at 50∶50, the response time of the EC devices prepared only by the WO3 film is the shortest. The coloring time and the fading response time are 2127 s and 11.68 s respectively. When the EC devices were using the Ag-WO3 composite film as the color-changing layer and the Ag layer thickness is 6 nm, the two response times are the shortest. The coloring time and the fading response time are 1271 s and 7.89 s respectively. The response time of the EC devices prepared by WO3 film is influenced on different argon-oxygen flow ratios. Moreover, the response time of the EC devices prepared by Ag-WO3 composite film is significantly shorter than that of the devices made by WO3 film. However, when the Ag layer thickness increases, the response time of the devices also increases.

In order to explore the influence of environmental humidity on the stability of the negative bias illumination (NBIS) of amorphous indium gallium zinc oxide thin film transistors (a-IGZO TFT), the paper employed an unsealed chamber to carry on TFT’s negative bias tress test under the illumination of lights of different wavelengths and different relative humidity. First, the basic structure of a-IGZO TFT and the I-V test system used in the experiment were introduced. Next, we tested the transfer characteristic curve of a-IGZO TFT under the illumination of light of different wavelengths and the transfer characteristic curve of different relative humidity under the illumination of the light of the same wavelength. The experimental results showed that the transfer characteristic curve of a-IGZO TFT shifted negatively as the voltage bias time increased. With the decrease of the wavelength of light, the device threshold voltage had a more significant negative drift. As the relative humidity increases, the NBIS instability of a-IGZO TFT gradually decreased but its electrical characteristics have seriously deteriorated. Environmental humidity has a significant impact on the stability of a-IGZO TFTs NBIS. When the relative humidity was 50% and the light wavelength was 400 nm, the threshold voltage shift of a-IGZO TFT reached 15 V while the negative voltage has stressed for 1 500 s, which showed the best performance under negative bias illumination stress.

Liquid crystal polarization grating is a new type of grating device based on the principle of geometric phase. It can modulate the phase and polarization state of incident light by adjusting the spatial distribution of the optical axis of liquid crystal. Traditional liquid crystal polarization gratings can achieve high diffraction efficiency at normal incidence and a small spectral bandwidth, but it is difficult to maintain a high diffraction efficiency with broad spectral bandwidth, and the diffraction efficiency decreases significantly when incident at a large angle. The proposed multi-layer twisted structure allows these problems to be solved, but introduces a new problem, how to accurately control the twist angle and thickness. Based on this problem, this paper proposes a new method for measuring the twist angle and thickness of a liquid crystal polarization grating based on Jones matrix. The difference fitting method eliminates the error caused by the refractive index change of the entire waveband, and can accurately obtain the twist angle and thickness of the liquid crystal cell. The similarities and differences between the twisted liquid crystal film and the twisted liquid crystal polarization grating are analyzed. Based on the measurement results of the twisted liquid crystal film, the twist angle and thickness of the liquid crystal polarization grating prepared under the same conditions can be obtained. The experimental results show that the thickness measurement error is less than 2%, and the twist angle measurement error is less than ±0.5°. This method can quickly and conveniently realize the accurate measurement of twist angle and thickness, with high measurement accuracy and stability. The relationship between the concentration of the chiral agent and the twist angle is analyzed, which provides a theoretical basis for the preparation method of multi-twist liquid crystal polarization grating.

There are many factors that affect the imaging of large-aperture ground-based adaptive optics system. In order to obtain high-resolution imaging near the diffraction limit, it is necessary to optimize the key parameters of the adaptive optics system. This paper analyzes the error sources that affect the imaging of ground-based optical telescopes and establishes a reasonable error budget model. The simulation method was used to verify the model, and optimized analysis of key parameters such as turbulence conditions, telescope aperture, sampling frequency, deformable mirror driver spacing, and guide stars. Compared with emulation, the results show that the prediction accuracy of the error model can reach within 30 nm under better observation conditions. This model also provides a reference for the application of adaptive optics in the fields of biological microscopy imaging, fundus imaging, and laser atmospheric communication.

Different construction orientations have a significant influence on the mechanical properties of the standard samples of fused deposition modeling (FDM). In this paper, polylactic acid (PLA) wire rod was used to prepare three kinds of printing samples in three different directions: standing, side standing and lying down, The digital image correlation (DIC) method was used to analyze and record the tensile and bending deformation process of the sample in real time. The real deformation and strain of the sample can be obtained effectively by using the measurement system, and the more accurate stress-strain curve can be obtained combined with the test machine data. Through the observation of the linear region in strain nephogram, it can accurately predict the time and location of fracture, and compare the mechanical properties of different oriented forming processes on printed products performance impact. The test results show that there is no obvious necking phenomenon in the tensile test and three-point bending test, and its deformation, elongation and strength are also the lowest, showing the characteristics of brittle materials. The lateral vertical has good tensile properties. After the necking phenomenon occurs, it will fracture after a small deformation, and the force that it can bear presents a linear downward trend until it breaks. The strength of the lying down type are similar to those of the side standing type, and the necking phenomenon occurs until the fracture deformation is larger than that of the side standing type, and the stress decreases gently with the increase of deformation, showing the best mechanical properties in strength and ductility.

The traditional 6T SRAM (5T and a driving MOS transistor) OLED-on-silicon pixel circuit is taken as an example to analyze the influence of switch size on data writing. A novel digital OLED-on-silicon pixel circuit is proposed, which uses 5T (4T and a driving MOS transistor) to achieve the same function as 6T SRAM (5T and a driving MOS transistor). In addition, compared with the traditional SRAM structure, the data writing of the new pixel circuit is not affected by the switch size, and the smallest switch size can be used. Based on SMIC 0.18 μm 1.8 V/5 V mixed signal process design, the simulation results show that the minimum size of the switch in 6T SRAM pixel circuit is 540 nm/600 nm, and the pixel unit layout area is 4.3 μm × 4.3 μm. The switch in new 5T pixel circuit can be the minimum size of the process, that is, 300 nm / 600 nm, and the layout area is 3.91 μm × 3.91 μm. In contrast, the area of a single pixel unit layout is reduced by 17.3%.

In order to satisfy the visual requirement of information sharing and confidential information protection, the research work mainly focuses on controllable visual angle technology applied to notebook display area, and proposes a new multi view switchable notebook display technology. The method combines backlight module with window-shades louver film, uses two cell design meaning consist of a dimming liquid crystal display (DLCD) device and a LCD display with micro lens, wide viewing mode and privacy mode can be switched by one keyboard toggle by modulating the signal of DLCD and the current of backlight. When users want to share information, it performs a sharing wide view mode. On this situation, the current of the backlight is higher and weak vertical electric field is applied to the DLCD. When it switches to a privacy mode for users to protect the display information, the current of the backlight switches to a lower state, and the vertical electric field of DLCD becomes strong. When the view angle is larger than 45° in horizontal direction, the display image is blocked well in narrow viewing angle mode. The wide view mode has the same wide view angle display characteristics as the industry wide view displays. The viewing controllable technology studied in this paper has a good application prospect in the field of anti-peep notebook display as the viewing angle can be switched freely with a high image quality and without sacrificed any aperture ratio.

The degradable optical alignment technology and its main effects on TFT-LCD are studied. The anisotropy of PI shows a trend from increasing to decreasing as the exposure dose increases and the main curing time extends. The anisotropy of PI, which is a common representation of the alignment ability of PI, performs best at 550 mJ/cm2 ultraviolet treatment and 30 min curing time. In addition, the anisotropy of PI keeps increasing as the main curing temperature increases from 230 ℃ to 250 ℃. Results also demonstrate that the complete optical alignment has no obvious effect on the electrical properties of TFTs. Compared to rubbing process, optical alignment shows huge advantages of contract ratio. The contract ratio of optical alignment product is about 1 500, approximately 50% higher than that of rubbing product, due to the low pre-title angle (about 0.2°).

Based on the causes and repair methods of particle gap, this paper analyzes the influence factors about the cell gap and head diameter for success rate. When the cell gap is different, the same pressure value has different success rate. The different cell gap has the differentiation of the optimal pressure value. When the cell gap is small, the pressure value required to achieve the best repair effect is larger, and vice versa. The change in the diameter of head directly affects the intensity of pressure acting on the TFT-LCD surface. With the intensity of pressure increasing, the success rate first increases and then decreases. The broken screen rate increases significantly after a certain pressure. The study shows that in order to ensure the repair effect of pressing particle gap in actual production, it is necessary to adjust the appropriate pressure for the different cell gap, and through the best intensity of pressure to select the appropriate diameter of head, to ensure a good repair effect.

In order to obtain better image super-resolution reconstruction quality and improve the stability of network training, the generation of confrontation networks and loss functions are studied. Firstly, SRGAN and DenseNet are introduced, a generation network is designed to generate image based on DenseNet, and the sub-pixel convolution module is added to DenseNet. Then, the redundant BN layer in the original DenseNet is removed to improve the training efficiency of the model. Finally, the loss function of SRGAN is introduced and the loss function is redesigned based on the Earth-Mover distance, and the SmoothL1 loss is used to replace the MSE loss to calculate the VGG feature map to prevent MSE from amplifying the gap between the maximum error and the minimum error. Experiments prove that the model can achieve a stable convergence state during the network training process. The quality of the reconstructed image is compared with SRGAN,the average PSNR on the three benchmark test sets SET5, SET14, and BSD100 is about 2.02 dB higher, and SSIM is about 0042 (5.6%) higher. The reconstructed image not only has improved indicators, but also has better definition and richer high-frequency details.

In order to meet the high requirements of the magnetic tile production industry for surface quality inspection and realize the automatic segmentation and recognition of magnetic tile defects, a defect segmentation and classification network based on convolutional neural networks is proposed. The network is based on the U-net architecture. The deep features of defects are extracted through the U-net encoding part, and the deep features are used for defect classification, and then the segmented defect areas are output through the decoding part. In order to solve the problem that the proportion of the foreground area of some defects is too small, which makes the network difficult to converge, the continuous optimization of the network is ensured by adding the difference coefficient loss. Then, adding multiple layers of loss and performing online data enhancement in the training phase further improves the segmentation accuracy and classification accuracy. Experimental results show that with the addition of auxiliary loss function and data enhancement, the segmentation network can segment 94.5% of the marked defect areas, and the accuracy of defect classification can reach 98.9%, which meets the high precision requirements of the industry. This method can accurately and effectively segment and identify the surface defects of the magnetic tile, which provides a new idea for the automatic industry of magnetic tile surface quality inspection.

In order to improve the accuracy and time efficiency of hippocampus multi-atlas segmentation, an algorithm based on Advanced Normalization Tools (ANTs) registration was proposed. In order to reduce the data size, a box with hippocampus as the center was extracted in the preprocessing stage. In the registration stage, ANTs were used to replace the resampling link, and the smoothness, topological retention and continuity of the differential Diffeomorphic Demons algorithm were used to perform accurate registration. In the tag fusion stage, four fusion algorithms including weighted average (Majority Voting, MV) algorithm, GraphCut tag fusion (Generative Model, GM) algorithm based on generated model constraints, metric learning (Metric Learning, ML) algorithm and semi-supervised tag propagation random forest (Integrating Semi-Supervised Label Propagation and Random Forests, RF-SSLP) algorithm were compared. The experimental results show that after replacing resampling with ANTs, the accuracy of four fusion algorithms including MV, GM, ML and RF-SSLP can be improved, respectively. Meanwhile, through the comparison of the above four fusion algorithms, it is found that the semi-supervised random forest algorithm based on ANTs registration has the highest segmentation accuracy, which is improved by 3%~5% compared with MV, GM and ML fusion algorithms.

In view of the difficulty of traditional methods for simultaneous and effective multi-target segmentation, the existing ground object classification methods based on fully convolutional neural networks have low classification accuracy in complex scenes, this paper proposes an improved encoder-decoder based on the U-shaped network structure DL-Unet, which realizes the effective segmentation of remote sensing images. This network improves the traditional convolution method and introduces the expanded convolution, which increases the receptive field without increasing the network parameters. Aiming at the problem of imbalance in the clssification of featares in remote sensing images, weighted cross-entropy is used as the loss function of the model, which effectively overcomes the selection preference of the model function of the model effectively. The relative majority voting strategy is adopted for the prediction results to further improve the pixel accuracy (PA) of each feature category. The experimental results show that the mean pixel accuracy (MPA) and mean intersection over union (MIoU) of this model are improved by 5.94% and 9.45% respectively compared with the classic U-net, which verifies that the method in this paper is an effective remote sensing image classification method.

Aiming at the problem of low accuracy of traditional image processing methods for liquid leakage detection in a factory with complex equipment structure, numerous types of debris, and severe ground wear, a leak detection algorithm based on CNN is proposed. The leak detection problems is analyzed, the data set is made, the VGG16 model is established. In order to avoid over-fitting state, combined with early stopping algorithm to train samples, the rapid and automatic detection of the leakage of complex pipelines is achieved. In industrial sites, this method can accurately identify the leakage and reduce the impact of noise interference. Finally, the superiority of this algorithm is verified by comparison with a variety of image processing methods. The results show that the test accuracy of the algorithm can reach 99.44%, and the prediction accuracy can reach 97.0%, which is higher than the accuracy of traditional image processing algorithms. The prediction time of a single picture is about 0.2 s, which can meet the detection needs of industrial sites.

The deepening of the number of network layers can weaken the ability to characterize the detailed information of the deep features of the flame target, and at the same time extract redundant features with low correlation, resulting in low flame recognition accuracy. Aiming at this problem, a flame detection method based on improved Faster R-CNN is proposed to improve the accuracy of flame recognition in deep networks. Firstly, the ResNet50 network is used to extract flame features, and the SENet module is added to reduce the redundant features of flame targets. Then, the deep features and shallow features are multi-scale feature fusion to enhance the detailed information of deep features. Finally, the network is trained to realize the recognition of flame targets positioning. In the experiment, the VOC flame data set is constructed for network training, the test set is used for detection, and the feature map visualization is compared. Compared with the model before the improvement, the AP value increases by 7.78%, the recall increases by 9.05%, and the precision increases by 12.54%. By combining the attention mechanism module and the multi-scale feature fusion mechanism, the flame target detection model proposed in this paper, can effectively extract the flame target feature, and the flame target detection result is more accurate.