Light field technology can boost image encryption technology from two-dimensional to three-dimensional, and enhance the security of encryption. The refocusing algorithm can be used to achieving image decryption. However, it will introduce interference between images. Based on the deep learning technology, the regularity of image interference is analyzed. The simulated light field data set is constructed. This paper creats a 7-layer full convolutional neural network. As for training the full convolutional neural network, the simulated light field data set is used as input, while the original images are used as labels and input into the full convolutional neural network. Then the real light field decrypted images are input to for testing. The experimental results show that the full convolutional neural network can decrease the interference of the optical field decrypted images obviously and improve the image quality effectively.

A method for joint detection of diseased leaves and healthy leaves and extraction of effective characteristic wavelengths for potato late blight was proposed. Principal component analysis was performed on the spectral images of healthy and late blight leaves, and the weight coefficient curves of principal component images were analyzed to extract six characteristic bands of healthy leaf and diseased leaf. Based on the six characteristic wavelengths, the second principal component analysis was performed, and the optimization was reduced to three characteristic wavelengths of 712.19, 749.70 and 841.47 nm. Based on these three characteristic wavelengths, principal component analysis was used to identify the diseased area with the most contrasting image of the main component, and the recognition rate was 100%. The characteristic wavelength of potato late blight disease could be achieved by combining healthy leaves with diseased leaves and secondary principal components. This technology provides a reference for the development of related equipment for potato leaf disease detection.

In view of the fact that digital images are prone to blur during processing, an adaptive deconvolution deblurring algorithm based on non-reference image quality evaluation is proposed. Firstly, according to the strong correlation between the no-reference image quality evaluation result and its distortion level, we can determine the fuzzy level of the image by calculating the no-reference image quality evaluation value and finally determine the convolution kernel with the linear relationship between the image fuzzy level and the fuzzy kernel. In order to ensure the fidelity of the color image before and after color processing and improve the efficiency of the algorithm, we propose to transform the distorted image color space to YUV, and only process the Y-channel in the distorted image. The Gibbs-like oscillation distribution phenomenon occurs in the neighborhood of sharp changes in the image gray levels. Gradient-based weight matrix is proposed to control the phenomenon. Experimental results show that the proposed algorithm can not only quickly and effectively remove the image blur, but also effectively retain the texture details of the restored image.

Long-term exposure of solar cells (photovoltaic, PV) to sunlight can cause the surface temperature to rise rapidly, thereby reducing its output voltage. In order to reduce the surface temperature of the solar panel, a thermoelectric cooler (TEC) on the back of the solar panel was studied to test the effect of the solar panel on the output performance of the solar panel at different room temperature. Studies have shown that when the TEC was not included and the ambient temperature was 26 ℃, 29 ℃and 33 ℃ and the maximum output voltage of the experiment was 2.32 V, 2.25 V and 2.20 V. After loading the TEC, the maximum output voltage at 26 ℃, 29 ℃ and 33 ℃ is 2.54 V, 2.59 V and 2.47 V. The output voltage is increased by 9.4%, 15.1%, and 12.3%, respectively. Therefore, loading a TEC in a solar panel can increase the output voltage and prevent the life of the solar panel from being lowered due to an increase in temperature.

Using the Zemax optical design software, a car reflector for small car headlights is designed. Aiming at the performance requirement of GB 7258—2017 Safety Technical Conditions for Motor Vehicle Operation, the optimization measures such as structural improvement are adopted for ordinary automobile reflectors to make the secondary light distribution effect more excellent, make the central spot more concentrated, and make the visual effect softer. Simulation results have shown that the effective working distance of the system is 150 m. The average light intensity of the central spot at 10 m is 4 552 lx, and the exit angle is from. 10° to +10°. The main performance indicators have reached the latest national standards, and have the potential to adapt to the future 3D printing mass production mode without increasing the industrial production cost.

Aiming at many current image reconstruction algorithms such as JPEG decompression and compressive sensing reconstruction, there are some problems such as unclear image and low resolution. This paper proposes a high-resolution image reconstruction mechanism based on code-decoding symmetric neural network. Firstly, the image is compressed to obtain a low-resolution image, and then the low-resolution image is used as an input image to encode-decode a symmetric neural network, and the convolutional neural network is used to encode the feature image, and finally the deconvolution neural network is used. Decoding implements detail recovery of the image. The experimental results show that the image reconstructed by the code-decoded symmetric neural network is clearer than the previous low-resolution image, indicating that the resolution of the image is improved.

The phase retrieval algorithm has always had some questions such as low precision, long convergence time and even stagnation. Based on the transport of intension equation (TIE)-Gerchberg-Saxton (GS) iteration hybrid algorithm improves the precision of phase retrieval, and gradient algorithm increases the iteration step and accelerates the convergence speed. This paper compares and analyses the GS-TIE algorithm and amplitude addition gradient algorithm retrieval effect from the perspective of simulation and experiment for GS-TIE algorithm and amplitude addition gradient algorithm. Based on the 2D images simulation for the comparison of two phase retrieval algorithms, it proves that the convergence speed of the algorithm is three times that of GS-TIE iteration algorithm, the accuracy of the algorithm is better than GS-TIE iteration algorithm, and the accuracy is ten times better than GS-TIE iteration algorithm. The experimental results show that GS-TIE algorithm retrieval phase is clearly visible, the contour is obvious, and it is excessively uniform at the edge, while the amplitude addition gradient algorithm is relatively fuzzy, and it is excessively uneven at the edge of the contour and has a large suspension.

The circular Airy beam is a type of autofocusing waves. The unique property of the circular Airy beam is that this beam can maintain quite low intensity profiles before the focal point and the maximum intensity of the beam can abruptly increase by orders of magnitude just at the focal point. The fine structures of the beam can be obtained by decreasing w. In this paper, firstly, the Fresnel, the Fresnel-Kirchhoff and the Rayleigh-Sommerfeld diffraction theories were compared. It found that the Rayleigh-Sommerfeld diffraction theory was the most suitable method. Then, the propagation characteristics of the circular Airy beam with fine structures were studied in detail by using the Rayleigh-Sommerfeld diffraction theory. The results showed that the nonparaxial components of the field would increase as w decreases. At beginning, the increasing nonparaxial components enhanced the autofocus, resulting in an enhancement of the peak intensity and a decreasing of the full width at half-maximum at the focus. However, as w further decrease and beyond a certain value, more and more nonparaxial components became evanescent wave, and could not propagate to far field. The peak intensity at the focus decreased sharply.

In order to realize more efficient and accurate automatic focusing of optical fiber end face, the definition evaluation function of existing images is studied. On the basis of this, a new definition evaluation function with simple calculation and good anti-noise is proposed. An optical fiber geometry measurement system is used to obtain 24 frames of equally spaced optical fiber end-face image. Different image definition evaluation functions are used to calculate the 24 frames of images. The corresponding focusing curves are obtained and their monomorphism, monotonicity and anti-noise are analyzed. The results show that the proposed evaluation function has better performance and has certain reference value for automatic focusing technology.

In order to improve the efficiency of modern manufacture and meet the requirement of 3D scanning, a scanning system which can accurately extract 3D contour information of object surface is designed. The 3D imaging method of structural light is adopted. Firstly, the grating is projected to the surface of the object under testing by ordinary white light, and the distorted raster image is collected by industrial camera. Secondly, the 3D coordinate information is extracted by Fourier transform contour method according to the change of gray value in the image of deformation raster. The experimental results show that the reconstruction of the three-dimensional point cloud with ideal effect can be achieved by using the Fourier transform contour method,which provides an experimental method for the 3D contour scanning and a reference basis for the precision of the high point cloud.

In the optical characterization of micro/nano materials, in order to obtain higher resolution wavelength and intensity fluorescence images, a common imaging microscope, spectrometer and nano mobile station are combined to form a laser scanning microscope imaging system. LabVIEW is used to develope a complete set of system PC software for integrating 2D scanning acquisition and signal image processing. The scanning acquisition process uses digital signal processing methods such as low-pass filtering to eliminate the effects of spectrometer signal noise. The system is used to measure the cadmium selenide nanobelts and monolayer molybdenum disulfide. The fluorescence intensity image and the fluorescence peak wavelength image are obtained. Fluorescence with a minimum wavelength of 0.03 nm can be resolved. The acquisition length is compared with the actual length and the fluorescence intensity difference is analyzed. Good results are obtained.

In order to detect optical signals in very low and high optical power range, a weak light detection system based on silicon photomultiplier (SiPM) is proposed. This system consists of auto-adjusting bias circuit, current-voltage conversion circuit, small signal amplification circuit and filter circuit. The relationship, between the output of SiPM and the incident light signal under different bias voltage was measured. The experimental results show that the bias has a great influence on the output of SiPM, the detection ability and detection range of SiPM under different bias voltage. In addition, the system has response of the 25 pW to 1.75 μW optical signals, which can achieve continuous detection of optical signals in very low and high optical power range.

The axial eye length is closely related to common ophthalmic diseases such as refractive error and cataract. Therefore, the accurate measurement of the axial eye length is of great significance for the prevention and treatment of ophthalmic diseases. Because the reflected light from the retina is just 10.3 -10.4 times of the incident light whose value cannot exceed the safety limit, axis eye length measurement based on optical method is a kind of weak signal detection. This paper designs a measurement system based on heterodyne interferometry. The system selects PIN photodiode to convert the weak light signal into current signal, designs a low-noise, high-performance weak signal processing circuit, and adopts the 10M25SAE of Max10 series as the main control chip to process signals as well. The experimental results show that the measurement system can achieve an effective and stable measurement of the axial eye length.

In order to enhance the control of the dynamic light source, the position and illuminance control system of the dynamic light source is designed by using the LabVIEW dynamic light source proportional-integral-differential (PID) control method. The system changes the position of the light source by controlling the motor, and changes the illumination of the dynamic light source by controlling the output current through the PID control and parameter self-tuning module of the programming software LabVIEW. The research shows that the system can effectively control the position and illumination of the dynamic light source, be applied to the detection of the dynamic range of the camera, and improve the detection efficiency and accuracy.

Laser interferometer plays an extremely important role in the discovery of gravitational wave. The Standard Quantum Limit (SQL) is one of the big limitation of the sensitivity improvement. The mechanisms and characteristics of squeezing light were deeply discussed. The techniques of SQL with squeezing light were briefly introduced.