A reasonable initial structure can effectively reduce the workload and improve the subsequent optimization efficiency in the design of aspheric system,especially the optical system with several aspheric surfaces.This paper proposes a method using simultaneous multiple surfaces (SMS) principle to construct the initial structure.The aberration characteristics of the method in imaging optical system design are analyzed,and the design process of initial structure of aspheric optical system is proposed based on the analysis.Finally,a laser radar receiving optical system with working wavelength 850 nm,focal length 20 mm,relative aperture 1/1.6,field of view 40°,and total length less than 30 mm is designed by the proposed method,which verifies the feasibility of this method in determining the initial structure of aspheric optical system.

An optical fiber plane strain sensor with droplet-like structure encapsulated by polydimethylsiloxane (PDMS) was proposed to monitor the deformation of thin slice workpiece.The sensor was formed by bending a piece of standard single mode fiber (SMF) to form a whispering wall structure,shaped like a droplet,and was encapsulated in a square mold made of PDMS.The magnitude and direction of the external strain on the workpiece were obtained by monitoring the change of the output spectrum wavelength of the sensor in the experiment.Experimental results showed that when the sensor was strained along the small axis of the droplet-like structure,the output spectrum was blue-shifted with the sensitivity of －0.108 nm/με.While the output spectrum was red-shifted when the sensor was strained along the large axis with the sensitivity of 0.084 nm/με.The experiment also studied the condition when the force direction was 45° with the short axis,which provided a reference for judging the strain direction,and temperature sensitivity of the sensor was －1.557 nm/℃.Therefore,the proposed optical fiber plane strain sensor could not only detect the magnitude of strain,but also distinguish the directions of strain through strain sensitivities,so as to achieve the purpose of plane strain measurement.The sensor will have a good application prospect in such fields as coating,bending and so on.

Based on the vernier effect of the parallel interference filter,we propose a fiber optic temperature sensor based on the paralleled structure of the fiber Sagnac interferometer (FSI) and the Mach-Zehnder interferometer (MZI).The FSI consists of a 3 dB four-port coupler and a 6.6 m polarization maintaining optical fiber (PMF).Because of its higher sensitivity and better stability,it is used as a sensing cavity.MZI as a reference cavity consists of two 3 dB three-port couplers self-made,by controlling the length of the two arms of MZI,so that the free spectral range (FSR) of the two interferometers is close but not equal,the vernier effect is used to improve the temperature sensitivity of the structure,and then by changing the temperature to measure the wavelength drift of a single FSI and paralleled FSI,MZI,so as to explore the amplification of temperature sensitivity.Experimental results show that the temperature sensitivity of a single FSI is only －1.65 nm/℃,and the paralleled system can amplify it to 12.9 nm/℃ with a gain coefficient of 7.82,which is consistent with the theoretical results,indicating that the paralleled structure can significantly improve the sensitivity of the temperature sensor at the same temperature.And the proposed sensor has significant wavelength drift while a smaller temperature changing which is suitable for fine temperature detection in biological and industrial fields.

Based on microwave photonic hybrid-packaging technology,a multichannel demultiplexed optical-to-electrical (O/E) conversion module is fabricated.Optical chip,microwave chip,free-space lens,and microwave circuit are packaged in a cavity,making the module combine the function of demultiplexing,O/E conversion and microwave processing to achieve the conversion from 1-channel multiplexed radio- over-fiber signal to 6-channel radio frequency (RF) signal and to output the signal after amplifying.The multichannel demultiplexed O/E conversion module is composed of several bare chips such as demultiplexer,focusing lens,photodetector (PD),microwave amplifier,equalizer,and power supply,which utilizing hybrid integrated packaging technology to greatly reduce the size of the module and improve integration and complexity of the module.The module shows huge potential in application of multi-functional microwave photonic area.The test results show that the module can realize the broadband conversion between radio-over-fiber signal to RF signal,and the efficiency of the O/E conversion exceeds 0.7 A/W.In addition,the test results of the amplitude-frequency characteristics show that the mean value of S21 is －8.4 dB,with a fluctuation of ±3.5 dB.The test results also show that the reflection of standing wave is less than 2 dB,and the isolation degree between adjacent channels is above 40 dBc.

Deeplabv3+ network does not make full use of multi-stage feature information in urban street view image segmentation,which leads to the shortcomings of large targets with holes,imprecise segmentation of edge target and so on.Considering the natural spatial position particularity of urban street view data,this paper proposes to introduce a height-driven effective attention model (HEAM) and a multi-stage feature fusion model (MFFM) on the basis of Deeplabv3+ network,and it embeds HEAM into the feature extraction network and atrous spatial pyramid pooling (ASPP) structure,which makes it pay attention to more spatial position information in the vertical direction.MFFM integrates multi-layer feature images to form multiple branches in the network and connect them to the network decoding end in turn.Successive up-sampling is used to improve the continuity of pixels during decoding.The improved network is verified and tested by CamVid urban street view data set.The results show that the network can effectively improve the deficiency of DeepLabv3+,and the location priori of the data set is properly used to enhance the segmentation effect.Mean intersection over union (MIoU) on CamVid test set reaches 68.2%.

To solve the problem of poor robustness and low accuracy of workpiece tracking in complex industrial production environment,this paper presents a multi-attention fusion workpiece tracking algorithm based on accurate tracking by overlap maximization (ATOM).The algorithm uses ResNet50 as the backbone network,first incorporating a multi-attention mechanism,which makes the network pay more attention to the key information of the target workpiece.Secondly,the attention feature fusion (AFF) module is used to fuse the deep and shallow features to better preserve the semantics and details of the target workpiece in order to adapt to the complex and changeable environment of industrial production.Finally,the third and fourth layers features of the backbone network are fed into the CSR-DCF classifier,and the resulting response graphs are fused to obtain rough locations of target workpieces and accurate target frames through the state estimation network.Experiments show that the Success and Precision of the algorithm on OTB-2015 dataset are 67.9% and 85.2%,respectively.The overall score on VOT-2018 dataset is 0.434,which has high accuracy and robustness.On the target workpiece sequence taken by the CCD industrial camera,the algorithm is further validated to meet the common challenges efficiently in the workpiece tracking process.

There is abnormal information about the fault in the bearing fault signal,which is significant to maintain mechanical safety.After the bearing fault signal is decomposed by wavelet packet,the abnormal information of the fault is mainly reflected in the dynamic error of the decomposed frequency band.The dynamic error of each frequency band is generally described by the energy entropy of standard deviation and the mean value of standard deviation.In order to highlight the distinguishing characteristics of bearing faults,the dynamic error is demarcated by the bearing fault size.It is an effective method to extract the relative dynamic error by using the corresponding bearing fault characteristic parameters.Based on this idea,this paper calculates the energy entropy and mean through the standard deviation of different frequency band components after wavelet packet decomposition.Then,the energy entropy of the standard deviation and the mean value of the standard deviation of the corresponding frequency band are added as the characteristic parameters for qualitative analysis at the same scale.At the same time,the value obtained by adding the characteristic parameters of the different frequency bands of the bearing signal is compared with the bearing fault size.Through the quantitative analysis of the relative dynamic error,the effective distinction of bearing faults is finally realized.The experimental results show that the method proposed in this paper has a good effect on distinguishing bearing faults.

F-norm is sensitive to outlier data,while L1-norm can significantly reduce the sensitivity and cannot effectively control reconstruction errors.To tackle the problem,we take both F-norm and L1-norm as the distance metric of the objective function,and propose a joint-norm two-dimensional principal component analysis (2DPCA) algorithm called 2DPCA-F-L1,and give its non-greedy solution.This algorithm not only ensure the ability of image classification,but also decrease the average reconstruction error in image reconstruction.When applied to underwater biometric image recognition,the proposed 2DPCA-F-L1 suppresses the noise interference in underwater optical images.Experiments show that the 2DPCA-F-L1 algorithm can accurately recognize the species of underwater creatures,and has better robustness than other principal component analysis (PCA) algorithms in image reconstruction experiments.

Based on the method of combining the extended Huygens-Fresnel principle with the Wigner distribution function (WDF),the analytical expression of M2 factor and angular width of partially coherent twisted vector beam (PCTVB) in ocean turbulence is derived.Numerical simulation shows that the anti-turbulence ability of beam can be improved to some extent by adjusting beam waist width,twisted factor,initial coherence length，wavelength.Decreasing the initial coherence length,increasing wavelength and waist width can make the beam have a smaller M2 factor,while increasing the absolute value of the twisted factor,the M2 factor of the beam is smaller,and the anti-turbulence interference of the beam is stronger.As the temperature variance dissipation rate and the ratio of temperature and salinity decrease,the dissipation rate of kinetic energy and anisotropy factor increase,the effect of ocean turbulence on beam becomes smaller and the beam will have better propagation quality.

Sources of single-photon in pure quantum states are vital resource for quantum information technologies,and have many important application values in new generation information technology,such as quantum communication and quantum computing.Here,we propose a scheme on the miniaturized and heralded generation of pure single-photon through periodically poled lithium niobate (PPLN) crystal.Under the conditions of group velocity matching and quasi phase matching,a non-degenerate spontaneous parametric down-conversion process is designed to generate frequency uncorrelated photon pairs,thus realizing the preparation of heralded pure state single-photon sources.The parameter conditions of frequency uncorrelation are deduced theoretically,and the yield of photon pairs under different conditions are calculated.It is found that the generation efficiency in the lithium niobate crystal waveguide structure are 4—6 orders of magnitude higher than that in the corresponding bulk crystal under the same parameters.These results will be helpful to improve the purity and the yield of miniaturized single-photon sources,and will play an important role in promoting the further development of integrated quantum optical chips.

A monocyclic nested hollow-core anti-resonant fiber (HC-ARF) with large transmission window and low confinement loss is proposed.The transmission characteristics of HC-ARF are simulated by full-vector finite element method combined with perfectly matched layer boundary conditions,and the influence of the structure parameters of HC-ARF on the transmission characteristics is analyzed.The simulation results show that the optimized fiber has the advantages of large transmission window,low confinement loss and flat dispersion.When the diameter of the fiber core is 50 μm,the number of the anti-resonant tube is 6,the thickness of the anti-resonant tube t is 0.30 μm,the diameter of the outer anti-resonant tube d is 32.50 μm,and the diameter of the inner anti-resonant tube d1 is 21.13 μm,the confinement loss is lower than 0.21 dB/km and the dispersion value is (1.1±0.3) ps/(nm·km) in the wavelength range of 1 260—1 675 nm,the confinement loss at 1 550 nm is 0.078 dB/km.

Medical image is often rich in pixels with the same intensity as the impulse noise so that medical image restoration in the presence of impulse noise is remarkably difficult.To gain a better capability of impulse noise reduction and structure preservation for medical image than the state-of-the-art filters in literatures,we propose a dual iterative equidistant mean filter (DIEMF) for medical image restoration.In the proposed method,an equidistant neighborhood is proposed for noise detection and removal processing;the noise detector discriminates noisy pixel from noise free ones by the averaged absolute difference between the neighboring non-extreme pixels and central pixel circularly,as well as majority rule;the noise removal technique uses adaptive and dual iterative method,takes the mean of noise free pixels and previous restored pixels in equidistant neighborhood as the estimated intensity of central noisy pixel,taking full advantage of the nearest previous restored pixels.Experimental results shown that the proposed method outperforms the state-of-the-art methods in noise reduction and structure preservation,especially for low noise level,it shows remarkable superiority over the state-of-the-art filters.

In nuclear medicine,single-photon emission computed tomography (SPECT) bone imaging is an important means to assist physicians in diagnosing diseases.Aiming at the problems of low signal-to-noise ratio,blurred boundaries,small lesions,and time-consuming manual lesion delineation in bone imaging images,an automatic segmentation algorithm for bone imaging lesions based on improved U-Net network was proposed.Based on the original convolution block of U-Net,the algorithm adopts a multi-scale dense connection (MDC) method to improve the extraction ability of small lesion features,and at the same time solves the problem of gradient disappearance after the network is deepened.Second,to extract detailed features of lesions,an attention mechanism structure is introduced at dense and skip connections.Finally,in view of the problem that the model is difficult to converge when using a small sample dataset,the transfer learning method is used to optimize the initial parameters of the model and improve the generalization ability and segmentation efficiency of the model.In addition,in order to reduce the amount of computation and further improve the segmentation effect,the dataset is cropped and denoised.At the same time,the processed images are augmented by rotation,mirroring and other methods.The experimental results show that the improved U-Net′s recognition precision and mean intersection-over-union ratio (mIoU) can reach 0.735 〖ＫＧ－１／６〗2 and 0.467 3,respectively,which are better than the current mainstream segmentation algorithms,and have certain practical application value.