The defect detection of photovoltaic modules based on temperature infrared image is an important technology to realize the large-scale modules quality detection of photovoltaic power plant. In this paper, the causes and hazards of hot spot of photovoltaic modules were briefly introduced. The artificial neural network model and its performance of infrared image and video of photovoltaic modules were summarized and compared from three aspects: hot spot detection, hot spot location and extraction. The hot spot detection accuracy of the improved YOLOv5 model for photovoltaic modules reached 98.8%, and the hot spot positioning accuracy of the Lucas-Kanade sparse optical flow algorithm reached 97.5%. At the end of this paper, the development trend of hot spot detection technology adapted to meet the operation and maintenance needs of large-scale photovoltaic power plant was briefly discussed.

A 1024 linear CMOS image sensor (CIS) with programmable resolution and high speed was presented and verified. In order to optimize the column bus readout rate, this sensor adopted column bus segmentation technology, which effectively reduced the parasitic capacitance on the bus and increased the signal readout rate. The sensor featured 4 kind of programmable resolution for higher frame rate. In this paper, the chip was successfully manufactured by 0.5μm CMOS process, and the photoelectric response of the device was tested and evaluated. The results show that the sensors full well charge is 4.76Me-/pixel, and dynamic range is 75dB. At 128 resolution, frame rate reaches 36000frames/s.

With the improvement of silicon material quality and the development of semiconductor process technology, the dark current of CCD (Charge-Coupled Device) under low-temperature refrigeration can be neglected. At this time, the amplifier becomes the main noise source, which limits the detection of weak signals. By using the electronic multiplication (EM) technology to amplify the charge packet before it enters the output stage, the noise of the amplifier can be effectively suppressed. However, the charge packets are easily saturated after amplification, which will limit the dynamic range of the device. In this study, the floating gate amplifier performed nondestructive detection on the charge packets. According to the detection results, the amplifier outputs the large charge packets directly, while the small charge packets are amplified by electron multiplication and then output, thus obtaining high sensitivity and large dynamic range. According to the actual test structure, good micro-light imaging is obtained under 10-4Lx illumination, while the dynamic range was up to 150,000 times.

In order to expand the measuring range of flow sensor and reduce power consumption of the sensor, a MEMS flow sensor based on self-heating amorphous germanium film thermistors was fabricated and tested. It was composed of four amorphous germanium thermistors and a pair of environmental temperature measurement compensation thermistors, which were embedded in a silicon nitride film. The four amorphous germanium thermistors acted as self-heating heat sources and temperature measuring elements at the same time, and were connected with each other to form a Wheatstone bridge. The MEMS fabrication process was given, and the size of 8.9mm×5.6mm×0.4mm of the flow sensor chip was fabricated by micromachining. The low velocity and high velocity airflow pipeline experimental devices were built. With constant current (CCA) of 50μA supplied to the Wheatstone bridge of the flow sensor, the flow velocity measurement of 0~50m/s was realized by test. The results show that the sensitivity of the sensor is about 81.6mV/(m/s) when measuring 0~2m/s low flow velocity, and about 51.9mV/(m/s) when measuring 2~50m/s high flow velocity, and the maximum power consumption is only about 1.03mW.

In this paper, three types of eight kinds of diode side-pumped structures were designed, and the corresponding theoretical model of pumping optics was constructed. The absorption efficiency and standard deviation of matrix elements under different structures were calculated by finite element analysis method. First, given the total pump power, chip spacing, doping concentration, etc., with the increase of the pump distance from 0~5mm, the absorption efficiency of laser crystal in the eight pump structures tended to decrease, and the standard deviation of matrix elements also decreased gradually. The standard deviation of matrix elements varied greatly when some structures were pumped at a close distance (0~0.3mm). In practical applications, the pump distance should be avoided in the above range. Secondly, under the same structural parameters, the absorption efficiency of the laser crystal was improved with the increase of the crystal doping concentration from 0.6% to 1.0%. At the same time, the increase of the doping concentration was the source of the reduced central peak of the cross section or light field depression in the radius direction. Therefore, the matching of doping concentration and pump distance will be an effective means to improve the uniformity distribution in the radius direction.

A low noise and low power phase locked loop (PLL) chip for the electronic readout system of high energy physics experiments was designed and tested based on the TSMC process of 180nm. The chip was mainly composed of frequency and phase discriminator, charge pump, loop filter, voltage controlled oscillator, frequency divider and other sub-modules. In the phase-locked loop charge pump module, the cascade current mirror was used to accurately mirror the current to reduce the electrical loss and the operation amplifier clamp voltage was used to further reduce the phase noise. The test results show that the PLL chip can stably output 200MHz differential clock signal under the condition of 1.8V power supply voltage and 50MHz reference clock input. The RMS clock jitter is 2.26ps (0.45mUI), and the phase noise is -105.83dBc/Hz at the frequency offset of 1MHz. The measured power consumption of the overall chip is 23.4mW and the core power consumption of the PLL is 2.02mW.

In recent years, with the wide application of photovoltaic detectors in the field of Fourier transform infrared spectroscopy (FTIR) detection, the nonlinear response of photovoltaic devices has increasingly become a key issue affecting the accuracy of FTIR spectroscopy inversion. Firstly, the general principle and equivalent circuit of the photovoltaic detector were summarized. Then, the internal effects and external factors leading to the nonlinear response of the detector were analyzed. Finally, the correlation between the internal and external factors was summarized, and the working direction of nonlinear correction of the detector was given in the conclusion. The nonlinear response of photovoltaic detectors was investigated in this paper. This is important for revealing the response law of semiconductor devices to different fluxes of radiation, for improving the study of nonlinearities in photovoltaic detectors, and for improving the accuracy of FTIR detection.

A general efficient charge to voltage factor (CVF) testing scheme is proposed by analysing the working principle of the photovoltaic conversion, charge transfer and charge output in CCD. In the scheme, the photosensitive zone of CCD transfered charge to the horizontal zone and the horizontal zone continuously output charge packets with a constant frequency by adding a DC bias on the photosensitive area of CCD and the continuously transfer driving sequential was added on the horizontal zone. Thus, the stable intensity response signal was output. Then, CVF was calculated by relationship between the reset leakage current and output signal intensity. According to the principle of the method, a general testing device for different CCD devices was designed to adapt to various CCDs, and a variety of CCDs were used for testing and verification. The result shows that the proposed scheme inproves the testing efficiency, accuracy and stability of the CVF.

A general mobility model that comprehensively considered the temperature, electric field strength and carrier concentration was adopted. The generation of carriers was calculated by using the actual solar spectrum and the absorption coefficient of non-fullerene materials. In combination with the drift diffusion equation and the current continuity equation, the theoretical modeling of high-efficiency organic solar cells was carried out. The current voltage curve, open circuit voltage light intensity curve and short circuit current light intensity curve of the device were calculated using this model. The results show that the current voltage curves calculated by the model are in good agreement with the experimental data, and the other two curves are also in good agreement with the experimental data. In addition, the effect of energy disorder on the device performance was analyzed by using this model. The results show that reducing the energy disorder of materials can improve the performance of organic solar cells.

With the development of wireless communication technology, miniaturized, multiband microstrip patch antenna has been widely used in wireless communication. In this paper, a hexagonal multiband microstrip antenna was designed with coaxial line feed and 120 degree slotting on the inside of each vertex of the hexagonal patch. Three different resonant modes could be effectively obtained by adjusting the geometry and size of the slotted patch, which could be applied to L-band and C-band. The simulation software CST was used to simulate and analyze this kind of antenna. The results show that the gain G≥6dbi and the return loss S11＜-10dB are all satisfied in the working band. The antenna has the characteristics of simple structure, small size and good gain, which can be applied to WiFi band in communication.

A fiber micro structure temperature strain dual parameter measurement sensor based on singlemode-multimode-singlemode (SMS) optical fiber interference structure cascaded fiber Bragg grating (FBG) is proposed. Its strain characteristics and temperature characteristics were experimentally studied. The multimode optical fiber with a length of 35.5mm was fused between two singlemode optical fibers by means of optical fiber fusion to form a singlemode-multimode-singlemode optical fiber interference structure, and a temperature strain dual parameter measurement sensor was made by cascading FBGs. The results show that, in the strain range of 200~2000με, the strain sensitivity of singlemode-multimode-singlemode interference structure and FBG is -2.31 and 1.22pm/με respectively, and the linearity is 0.9992 and 0.9994 respectively. In the temperature range of 580~700℃, the temperature sensitivity of singlemode-multimode-singlemode interference structure and FBG is 58.79 and 13.64pm/℃ respectively, and the linearity reaches 0.9967 and 0.9982 respectively, which can realize simultaneous measurement of temperature and strain parameters.

We have measured the temperature dependence photoluminescence (PL) of quantum dots (QDs) with/without AlAs cap grown on (113)B and (100) GaAs substrates by molecular beam epitaxy. The carrier thermal transfer was investigated by analyzing the integrated intensity, peak energy, and FWHM of PL spectra. For QDs grown on (113)B GaAs without AlAs cap, PL quenching can be explained as carriers are easily escape from QDs to WL. However, for QD grown on (113)B GaAs with AlAs, the carriers might escape out of the QDs into non-radiation centres in the QD-WL layer or QD-barrier interface. Temperature dependence PL peak energy of QD close to the temperature dependence of bulk InAs band gap, suggesting that the carriers transfer through WL has been inhibited since the WL was removed or reduced due to phase separation by AlAs cap. For QDs with AlAs cap grown on (100) GaAs, similar FWHM dependence to QD without AlAs cap were observed, indicating that the reduction of WL due to phase separation by AlAs might not as effective as that of (113)B GaAs.

Antireflection film with a narrow range of angle of incidence (AOI), antireflection film with a wide range of AOI, and multilayer high reflection thin film at 193nm were designed respectively. The maximum deviations of the s and p polarized light transmittances of the antireflection film were 0.17% and 0.44%, respectively. Combining the scalar scattering theory and the equivalent absorption layer approximation theory, the rough interface between the multilayer film was equivalent to a thin absorption layer, and the spectral properties at different interface roughness were analyzed based on the thin film intrinsic transfer matrix. The results show that the spectral performance of the optical thin film decrease with the increase of the root mean square roughness of the interface and the reflection bandwidth of the high reflection film also decrease. When the root mean square roughness reaches 4nm, the spectral performance of the antireflection film with a wide range of AOI and the multilayer high reflection film degrade by 2.04% and 2.09%, respectively. Interface roughness is the key factor affecting the preparation of VUV optical thin film with high spectral properties.

Mn2+ ion doped all-inorganic perovskite (Mn2+∶CsPbX3, X=Cl,Br) nanocrystals have been promising for a wide range of applications in solid-state lighting, photoelectric detection and imaging due to their advantages of wide emission bands, long Stokes shifts and high quantum yields. However, the quantum yield of Mn2+-doped perovskite crystals hardly exceeds 70%. How to improve the luminous efficiency while regulating the Mn ion emission centers becomes the key to build high-quality white LEDs. In this paper, highly efficient luminescent Cd2+ and Mn2+ co-doped CsPbCl3 nanocrystals were obtained by CdCl2 post-treatment technique with cation exchange at room temperature. The emission wavelength of Mn can be continuously tuned from 604nm to 624nm to achieve stable red light emission. Cd ion doping improves the crystal field environment of Mn-Cl octahedra and increases the Mn decay lifetime to 1.32ms. In addition, warm white light-emitting diodes (WLEDs) with high color rendering index were constructed by green CsPbBr3 nanocrystals and blue-orange bicolor Mn∶CsPb(ClBr)3 nanocrystals. The lumen efficiency reaches 60lm/W and the color rendering index exceeds 85.

Laser annealing is an important process of eliminating the backside well of the backside-illuminated charge-coupled devices (CCD) image sensors. The influences of different laser wavelengths, energy densities, overlap rate of laser spots on doping activation rate, device surface morphology, imaging quality and ultraviolet (UV) quantum efficiency of CCD were investigated. The results show that, the doping activation rate of 355nm laser is higher than that of 532nm laser at shallow junction implantation. However, the presence of wafer crack is easier to appear at low energy density when using 355nm laser. By adopting energy density of 2J/cm2 and 50%~65% cross rate of 355nm laser, the boron doping is efficiently activated, the imaging uniformity of backside-illuminated CCD is achieved, and the UV quantum efficiency of CCD is significantly improved.

The monitoring of radio frequency signals has broad application prospects. A microwave photon Radio Frequency (RF) signal monitoring system based on DSP is proposed. The system consisted of the front-end microwave photonic signal receiving part and the back-end signal processing part. The RF signal was processed by optical down-conversion to get the Intermediate Frequency (IF) signal. After the Analog-to-Digital conversion, the FFT algorithm was applied on DSP to complete the spectrum analysis of the received signal. The experimental results show that the system can complete the RF signal monitoring. The frequency measurement error is less than 0.25MHz, the dynamic range is 55dB, and the sensitivity is -30dBm. In addition, for further verification, the antenna was used to receive the mobile phone signal to realize the real-time monitoring of the mobile phone signal communication frequency band.

In recent years, the autonomous flight of bionic flapping wing aircraft using vision systems has become a widely promising research direction. However, its limited carrying capacity puts forward strict requirements for the type, size and weight of vision sensors. At present, the size and weight of commercial image processing modules are large, and the image information needs to be transmitted back to the ground control system for processing. Therefore, this paper aims to design a lightweight onboard monocular vision system to help the miniature bionic flapping wing aircraft obtain external information and realize intelligent and autonomous flight. Compared with other image processing modules, this system was designed with domestic high-computing power chip K210 as the core, which could complete image processing without the computer side. The size is only 2.2cm×2.3cm, and the weight is only 3g. The internal compatible lightweight network model realizes classification recognition, and information interaction is achieved through the serial port, which can control the flapping wing aircraft to achieve gesture recognition and target tracking.

In view of the large similarity between the weight limit sign and the speed limit sign, which is easy to cause false detection and other problems, combining the two-dimensional OTSU automatic threshold segmentation method with strong anti-noise ability and the Jumping Spider Optimization Algorithm (JSOA) with strong searching ability, a two-dimensional OTSU weight limit sign recognition algorithm based on improved JSOA is proposesd. The algorithm counted the two-dimensional gray histogram of region of interest of traffic signs, solved the threshold of segmentation of two-dimensional OTSU using the improved jumping spider algorithm, quickly realized the binarization of traffic sign images, used DBSCAN to cluster the edge points of the binarization image, and finally correctly identified the weight limit sign according to the relative position of the clustering results. The experimental results show that the operation time is reduced by 34.16% without reducing the recognition rate, and the weight limit sign and speed limit sign can be correctly distinguished.

In order to realize non-cooperative target recognition and tracking under weak echo signal, a target recognition and tracking strategy based on single pixel photon counting lidar system was designed, and a target recognition and tracking method was proposed. In this method, first, the 3D point cloud collected by the single-pixel photon counting lidar system was interpolated to obtain the intuitive range image. Then the maximum stable extreme value region algorithm was used to segment the target and background, and the target to be tracked was identified and selected according to the target contour features. Finally, the centroid of the identified target was extracted, and the deviation between the centroid and the scanning center of lidar was used as the error signal, and the servo system was controlled to perform target tracking on the basis of scanning. The experimental results show that when the laser emission energy is 625pJ and the number of echo photons is 25, the system can stably recognize and track the target with a distance of 5m and an angular velocity of about 2mrad/s. It is verified that the strategy and method based on single-pixel photon counting lidar can stably segment the target and background, and correctly extract the target centroid that needs to be recognized, tracked and ranged. It provides an effective and visual detection method for target recognition, tracking and ranging under weak echo signal. Meanwhile, weak signal detection is a necessary condition for long-distance detection. It provides a new technology direction for target detection and recognition tracking at long distance.

In order to solve the problems in simultaneous localization and mapping (SLAM), such as insufficient localization accuracy, accumulation of error of matching feature points and long matching time of feature points, a fusing improved RANSAC optical flow optimization algorithm is proposed. Based on the traditional RANSAC algorithm, the least square method was added to iteratively optimize the model to estimate the optimal model, and the mismatching points of optical flow method were removed to reduce a large number of image mismatching feature points. Then the improved RANSAC optical flow method was fused with the feature points through Kalman filtering. Finally, the improved algorithm was used to perform SLAM localization accuracy experiments in the open EuRoC MAV data set. Experimental results show that the improved algorithm in this paper can effectively reduce the feature matching error of optical flow method, thus improving the positioning accuracy of UAV visual SLAM, which proves the effectiveness and feasibility of the improved algorithm.

The spectral calibration of the single-lens imaging spectrometer was studied. The relative relationship between the wavelength and the image plane was marked by the frequency sweep of the monochromator, and the characteristic spectral line of the mercury lamp was used for verification. According to the selected single lens model, the calibration scheme and steps were designed. An experimental platform was built according to the scheme, and the experimental analysis and verification were completed. The final results show that the calibration method in this paper is suitable for single-lens imaging spectral equipment.

Power is one of the basic parameters of microwave signal. Traditional power measurement methods are difficult to adapt to distributed long-distance network applications due to high transmission loss and large volume of coaxial cables. In this paper, a method of microwave power measurement based on electro-optic conversion is proposed. The microwave power measurement was converted to the measurement of optical sideband rejection ratio. The mapping relationship between microwave power and optical sideband rejection ratio was analyzed respectively for electro-optic conversion methods using phase modulation and intensity modulation. Finally, the compliance between the measured value and the theoretical value was compared. After electro-optic conversion, this method can use optical fiber to transmit signals and carry out long-distance testing, which has a great application prospect in distributed long-distance networking.

A joint power and sub-channel resource allocation algorithm was designed for orthogonal frequency division multiplexing (OFDM) based multi-colour visible light communication (VLC) systems. Compared to the conventional monochromatic optical communication system, this algorithm could be used for OFDM signal transmission of multiple visible lights including red, green and blue. The proposed algorithm aimed to maximize the transmission rate of the visible light transmission network. The optimally allocation of sub-channels was first performed for three different visible lights. After obtaining this optimized sub-channel value, the power was allocatied to it in order to achieve the final target optimized value. Simulation results show that the algorithm can achieve higher transmission rate than fixed resource allocation algorithms.

Partial discharge (PD) monitoring is an important way to find the insulation defects of power equipment and maintain the safe operation of power grid. Based on the advantages of Distributed Feedback (DFB) laser, such as narrow line width, low noise and long coherence length, a vibration sensing system for partial discharge monitoring was designed. The DFB laser glued on the cantilever beam was used as the sensing head, the wavelength change of DFB laser was converted into the output intensity change of the interferometer by using the fiber Michelson interferometer, and the Phase Generated Carrier (PGC) algorithm was used to demodulate partial discharge signals. Cantilever was designed and vibration characteristics were simulated, and its first order resonant frequency was 876Hz. The relationship between the arm length difference of Michelson interferometer and the system sensitivity was studied. The pulse igniter was used as the partial discharge source for discharge detection, which verified the feasibility of the system for partial discharge detection. The results show that the system can detect the discharge signal of 800~900Hz with the sensitivity of -74.67dB re rad/Pa.

Endmember extraction is a key step in the decomposition of mixed pixels in hyperspectral remote sensing images. The traditional linear endmember extraction method ignores the nonlinear mixing factors of the geological features in the pixel, which restricts the improvement of the decomposition accuracy of the mixed pixel. Considering the nonlinear structure of hyperspectral image data, an orthogonal projection endmember extraction algorithm based on geodesic distance (OPEEGD) is proposed. The geodesic distance was introduced into the endmember extraction process, and the endmembers were extracted one by one using the orthogonal projection method. In order to reduce the amount of geodesic distance calculation, the original hyperspectral data was reduced by the automatic target generation process method and the unconstrained least squares method before endmember extraction. Experiments on simulated and real hyperspectral remote sensing images show that this method can characterize nonlinear factors in spectral data, and the endmember extraction results are better than traditional automatic target generation process endmember extraction method.

Aiming at the noisy signals collected by distributed optical fiber sensing systems, a noise reduction method to modified ensemble local mean decomposition (MELMD) combined with independent component analysis (ICA) is proposed. The permutation entropy decision mechanism was introduced to improve the ability to suppress modal aliasing and spurious components. Firstly, the product function (PF) was obtained by decomposing the noisy signal by MELMD method, and the signal reconstruction was carried out. The difference between the noisy signal and the reconstructed signal was calculated to obtain the virtual noise, and the virtual channel was constructed. Then ICA was used to separate the signal-to-noise of the noisy signal and the virtual channel to obtain the final result. Experimental results show that the proposed method can better eliminate the noise in the signal, and retain the characteristic information of the signal.

A chip open/short circuit defects inspection method is proposed based on infrared images series. Firstly, the mean temperature series of the critical area of chip during the response process of the power-on procedure was recorded, the Savitzky Golay convolution smoothing method was applied to extract the time domain feature parameters after smoothing and filtering, and the principal component analysis method was uesd to select the key features. Then a support vector machine classification model was constructed, whose parameters were optimized by particle swarm algorithm to effectively distinguish different types of circuit board defects. In order to prove the validness of the method proposed, a variety of solder ball open/short circuit experiments were carried out on a CPU chip of circuit board. The research results show that the cross-validation classification accuracy of the optimized SVM model in the test dataset is 96.90%, which proves the validness of the method for detecting the chip open/short defects in this paper.