The metal oxide thin film transistor (MOTFT) has the advantages of high carrier mobility (10~100 cm2/V·s), good thin film uniformity, low preparation temperature and transparent visible light. It is the most promising new generation TFT. Laser annealing has the characteristics of high energy, short time, low damage to the substrate and controllable processing range. According to the traditional thermal continuous, it is more suitable to produce flexible and large-size panels. The research progress of laser annealing discusses in detail the key parameters that affect the inversion effect in laser annealing; the effect of laser on metal oxide thin film and the effect of laser on the performance of MOTFT are systematically decomposed. Finally, the problems and development directions of current laser annealing technology are summarized.

With the demand on increasing data storage capacity and information processing speed in the modern society, more compact and integrated planar optical devices have received widespread attention. As one of the typical liquid crystal (LC) devices to modulate geometric phase, LC polarization gratings have shown great application potential in optical systems due to the advantages such as high diffraction efficiency, large deflection angle, and short response time, etc. However, some of the problems related to electro-optical performances have not yet been overcome in the current manufacturing process of LC polarization gratings, limiting its practical application. The control of diffraction characteristics of LC polarization gratings is subject to certain preconditions. In this work, by optimizing the elastic continuum model used to analyze the LC polarization grating, the influence of LC physical parameters (elastic constant, dielectric constant, etc.) on the critical cell thickness and threshold voltage, under the strong and weak anchoring conditions were investigated, providing the theoretical support for the selection of LC materials with suitable physical parameters for LC polarization gratings toward different applications.

The columnar defect of hole etching in oxide process was investigated in this study. The reason or sources which is responsible for the columnar defect was revealed and corresponding measures for avoiding columnar defect were effectively established．The primary reason for producing columnar defect might be related to film structure, which can be deduced from the differences between two kinds of hole. After uncovering each layer of the film, columnar defect was detected in ITO layer. The result of scanning electron microscopy (SEM) showed that small particles appeared on the ITO layer after hole etching. Based on the above-mentioned mechanism of columnar defect formation, three solutions were attempted to solve the issue. ITO crystal cannot be etched by etching solution and residual ITO crystals will impede SiO2 etching, which will produce columnar defect. Once ITO crystal was formed, extending the etching time will not remove the columnar defect. Decreasing the thickness of ITO can partially eliminate the columnar defect while completely erasing the columnar defect and etching the ITO crystal can be achieved by increasing the etching gases in passivation layer etching process

In order to improve cell gap stability of crystal display products, the relationship between the elastic recovery rate (ER) of photo spacer (PS) and the substrate films in thin film transistor liquid crystal display (TFT-LCD) was studied in this work. Firstly, PS samples with different substrate films were obtained by photolithograph process. Then, the SNU 3D microscope was utilized to characterize the size and height of PS samples, and the thicknesses of substrate films were measured by SEM. The ER was measured by dynamic ultra micro hardness tester (DUH) at last. Test results show that the ER of PS on bare glass is 3.9% higher than on normal substrate films when the loading force is 200 mN. While the Vickers hardness of over coat (OC) film vises, the ER of PS enhances obviously especially with higher loading force. The discrepancy of material A and B increases to 6.3% with 500 mN loading force. It is also found that the ER of PS decreases with the thicknsess of the substrate film of color resist. When the color resists thickness changes from 2.2 μm to 2.5 μm, the ER of PS on A and B decreases from 3.7% to 4.9%. In addition, the fatigue properties of PS on its base coating films are also studied. The results show that the elastic recovery rate of PS increases with the proportion of plastic deformation decreasing, and keeps steadily after 25 times as the number of press-in increases.

To reduce the energy consumption of LED backlight LCD, the four-primary-color backlight schemes are proposed. The energy-saving performance and chromatic aberration of the four-primary-color LED reproducing colors are studied. Firstly, since the human eye is not sensitive to red light, a method to use the fourth primary color instead of red is proposed to represent the color. To achieve the best energy-saving effect, the radiate luminous efficiency is introduced to determine the fourth primary color. Next, the method of converting the three primary colors to the four primary colors is explained by means of color gamut partitioning. Finally, how to evaluate the chromatic aberration of the reproduced colors is introduced. The results show that by using OGB colors instead of the popular RGB colors to reproduce composite colors, the average power consumption is reduced by more than 20%, and the average chromatic aberration of the reproduced colors of OGB and RGB is less than 1.5. This paper proposes a four-primary color LED backlight scheme, which has a good energy-saving effect for LCD and meets the requirements of current display accuracy and energy saving.

With the continuous development of display technology, the comprehensive screen display has become highly sought-after by the market. In order to minimize the display border and improve the screen share of display products, a new technology of laser curing anisotropic conductive film (ACF） is studied to achieve the display bonding. Through the tests of TFT electrode indentation, particle deformation, ACF reaction rate, ACF die shear and product reliabilities, it is found that laser bonding is a viable bonding technology. In this paper, the experiments are designed to verify the advantages of laser bonding technology. Conclusions are as follows: (1）Compared with hotbar bonding technology, the safety distance of thermal is reduced from 0.75 mm to 0.15 mm, which greatly shortens the IC/COF to polarizer(POL） distance and avoids the shrinkage of POL scald. (2）Because of its concentrated laser beam and directly ACF cured, the main bonding time is reduced from 5 s to 3 s, which greatly speeds up the production takt time and improves the efficiency. All results show that laser bonding is a very promising new technology.

In order to improve the characteristics of TFT, 4-mask PR ashing process was carried out by indu cti vely coupled plasma etcher. After PR strip, linear a-Si film residue appeared on the edge of source/drain data line,TFT channel and other pixel area. The effect of PR ashing conditions on a-Si film residue was studied. The results show that the pressure and Bias RF power are the main factors to produce the film residue, and the amount of O2 is the secondary factor. By adjusting the PR ashing recipe, the optimal condition for preventing film residue is obtained: Pressure ≥ 2.66 Pa, Source power∶Bias power ≥ 3∶1, qv(SF6)∶qv(O2) ≥ 1∶60. The research of inductive coupling plasma etcher in PR ashing process has very important significance for the further application.

Defect detection plays an important role in yield improvement for TFT array process. Traditional manual recognition is inefficient, while the emerging convolutional neural network for target detection needs a lot of manpower in defect labeling. In order to realize the automatic detection of TFT array defects while reducing labor costs as much as possible, a TFT array defect detection method based on Generative Adversarial Networks and morphological reconstruction is proposed. In this method, the dataset used to train the network does not need to be manually labeled, which solves the problem of high manual labeling costs. This method first obtains the attention mask of the TFT array through the AttentionGAN network, secondly, selects the least significant pixel in the attention mask as the seed point, obtains the defect mark image and the defect mask image, and then performs the region growth for binary morphological reconstruction, finally get the bounding box of defect. This method can achieve an F1 score of 0.94 for the two-class classification of TFT array defects, which proposes a new idea for automatic defect detection of TFT array.

In order to improve the privacy protection of personal information for electronic devices, the display devices are required to have the function of switching between wide and narrow view angle, especially in many public places. In this paper, a four-ways view angle controllable liquid crystal display technique is investigated with a specially designed pixel by dividing a sub-pixel into two domains, which realizes the wide and four-ways narrow view angle switchable with a single liquid crystal alignment direction. The display device described in this paper includes a view angle controllable electrode in the top substrate, the pixel and common electrodes are placed in the bottom substrate. When the view angle controllable electrode is set as common voltage, the fringe field is formed between pixel and common electrodes make the liquid crystal rotate horizontally toward different directions, so the wide view angle mode is realized. When the alternating voltage is applied to the view angle controllable electrode, the liquid crystal molecules rotate and have a large tilted angle under the action of horizontal and vertical electric fields, the liquid crystal molecules have a large phase delay at off-axis direction, and induce the light leakage in the dark state, so the four-ways narrow view angle effect is achieved. The experiment results show that the wide view angle can reach more than 60° in vertical and horizontal direction and the pictures are clearly visible. In the narrow view angle mode, it is perfect blocked in the view angle at 60° of polar angle in vertical direction and 45° in horizontal direction. The technical structure is simple, without any extra components, and the four-ways view angle switchable structure can be realized with one liquid crystal cell and one backlight.

In order to solve the problem of unstable training and low recognition accuracy of traditional deep learning algorithm in handwritten digit recognition when training samples are insufficient, a recognition method based on conditional generative adversarial network is proposed. Firstly, on the basis of the conditional generative adversarial network, the generator of category labels is used to control image generation. The image samples generated by the generator are used to expand the training data set. At the same time, the deconvolution network and the convolutional network are used to construct the network structure of generator and discriminator respectively, and the full connection layer is removed to improve the stability of the model. Then, the conditional batch normalization is introduced. Taking advantage of its use of category labels to make the network learn more features. Finally, the discriminator is improved to be a classifier, and a novel loss function is proposed to speed up the convergence rate and improve the recognition accuracy. The experimental results show that the handwritten digit recognition method proposed in this paper has better image quality and higher recognition accuracy of 99.43%. This paper provides a reference for the application of generative adversarial network and its variants in the field of image recognition.

In view of the complex background of the instrument recognition system, insensitive to small targets, and low detection accuracy, an automatic instrument identification method combining Feature Fusion Pyramid (FPN) and Faster R-CNN network is proposed in this paper. First, FPN and the RPN of the Faster R-CNN network are used to combine the positioning of the dial and the pointer area, and to classify multiple types of meters. In addition, in order to balance the positive and negative samples of the meter images and improve the detection accuracy, the Focal Loss loss function is combined with the RPN network to train the data set. The FPN-based image segmentation is performed on the pointer area. The FPN network and deconvolution are combined to improve the accuracy of the pointer area segmentation, and finally the pointer is fit to obtain the pointer deflection angle and obtain the meter reading.The experimental results show that the accuracy of the proposed method reaches 94.25%. Compared with the traditional algorithm, the proposed method not only has high detection accuracy, but also has stronger practicability.

In order to solve the problem of low face recognition rate in controlled scene, an expression recognition algorithm based on residual rectification enhanced convolutional neural network is proposed. This method takes convolutional neural network as the prototype. In the process of training model, the idea of residual network is introduced to correct the difference between the effect of test set and the effect of training set. The linear rectification operation of the residual block by the excitation function embedded in the convolution layer helps to express complex features. At the same time, the data enhancement method is used to suppress the fast fitting of the deep neural network model during the training process, and also to improve its generalization performance on a given recognition task, and then to improve the robustness of the model learning effect. In the experiment, the method is applied to simulate the online teaching environment, and the effect of effective facial expression recognition in controlled scene is achieved. According to the experimental data, this method can effectively classify the facial image input under controlled conditions, and the highest accuracy is up to 91.7%. This research is helpful to the development of facial expression recognition and human-computer interaction.

For the point cloud registration problem in the vision measurement, a point cloud registration method based on electro-magnetic tracking system is proposed in this paper. First, the measurement and registration systems are built, and coordinate system transformation relation between world coordinate system and transmitter coordinate system is calculated using coordinate system transformation method based on the flat surface. Second, the accurate position and orientation of the receiver relative to the transmitter are used to calculate the transformation relation of transmitter coordinate system and receiver coordinate system. Finally, transmitter coordinate system is taken as intermediate coordinate system. The point cloud from different view can be transformed from world coordinate system to receiver coordinate system, and the point cloud coarse registration is accomplished, then the Iterative Closest Point (ICP) method is used to conduct the fine registration. The registration experiment is conducted using the flat surface with different depth, and the plaster model is also used to verify the registration vision effect. The experiment results show that the registration error of the flat surface is below 0.40 mm after the coarse registration, and the registration error of the flat surface is below 0.04 mm after the fine registration. After the coarse registration, the error of the plaster model can converge quickly when we conduct the fine registration using the ICP method, and the vision effect is good.

Aiming at the problem that machine vision is difficult to detect defects on reflective surfaces, a defect detection method based on phase measuring deflectometry is proposed in this paper. This method uses an LCD screen as a structured light source, Gray code pattern combined with phase shift pattern as structured light, which is projected onto the measured surface through the LCD screen, and monocular camera is used to capture the reflected pattern of the structured light. Firstly, the image captured by the camera is decoded pixel by pixel to obtain the wrapped phase and fringe order. Secondly, the absolute phase image is generated by unwrapping phase and eliminating the error. Thirdly, Canny operator is used to detect the edge of the image to segment the projected region and the unprojected region. Finally, Canny operator is used to detect the edge of defect in the projected region to realize the detection and location of surface defect. The experimental results show that under the condition of low and constant ambient brightness, phase measuring deflectometry can efficiently detect the defects with deformation characteristics on the reflecting surface, with an accuracy of 0.1 mm. It basically meets the requirements of high precision, high efficiency and low cost detection for various defects on the reflected surface.