The influence of the tilt angle of the intracavity etalon on the output longitudinal mode distribution of Nd:YVO4 multi longitudinal mode laser was studied by rotating the linear laser intracavity gain medium. A theoretical model was established to analyze the effect of etalon effect on the output multi longitudinal mode characteristics of Nd:YVO4 laser, and corresponding experimental research was conducted. In the experiment, longitudinal mode grouping and spectral drift phenomena were explored by rotating the crystal angle. The experimental results indicate that when the laser crystal angle increases from 0°, there are two phenomena: the broadening of the transmission peak linewidth of the etalon and the redshift of the longitudinal mode wavelength. The experimental results are in good agreement with the theoretical model. Finally, when the pump power was 2 768 mW and the laser crystal tilt angle was 0.45°, the etalon effect was successfully suppressed, and a mode-locked laser with slope efficiency and optical conversion efficiency of 15.66% and 13.9%, output power of 385 mW, and beam quality of 1.71 was obtained. At this point, by increasing the pump power, it was found that the spatial hole burning effect also produces an effect similar to the etalon effect. The experimental results also indicate that increasing the pump power will exacerbate the phenomenon of spatial hole burning.

Optical fibers play an important role in terahertz (THz) wave transmission systems, however, there is a lack of transparent materials used to fabricate THz fibers, which limits the transmission distance of THz fibers. To further improve the transmission characteristics of THz fiber, a high birefringence hollow-core anti-resonant fiber is designed. The fiber cladding consists of a circular dielectric tube and a rectangular dielectric layer. The fiber structure is asymmetric. The finite element method is used to analyze the influence of fiber structure parameters on its loss and birefringence characteristics. The results show that the birefringence can reach 3.75×10-4, and the loss of x and y polarization modes are 0.077 dB/m and 0.395 dB/m at 0.475 THz, respectively. When the x polarization mode loss is ensured less than 0.1 dB/m, the high birefringence characteristics of THz fiber are achieved.

Accurate detection of surface defects on aircraft blades is crucial for ensuring the safe and reliable operation of aero-engines. Currently, vision-based algorithms for detecting surface defects on aircraft blades suffer from poor real-time performance, high missed detection rates, and inaccurate target localization. To address these issues, this paper proposes an aircraft blade surface defect detection algorithm based on deep neural networks. To improve detection real-time performance, we design the depthwise separable convolution (DSC) model to decompose standard convolutions. To reduce missed detection of small defect targets, we propose the squeeze-and-excitation path aggregation network (SE-PAN) model to recalibrate the features of each channel, allowing features with stronger information to receive more attention. To enhance localization accuracy, we design the focal-distance intersection over union (Focal-DIOU) loss function to mitigate the effect of inefficient boxes. Experimental results on our aircraft blade surface defect dataset demonstrate that our algorithm achieves Precision, Recall and AP of 95.7%, 94.6% and 96.3%, respectively, with a detection frame rate of 24 frames per second, all of which outperform mainstream detection algorithms.

In order to solve the problems of insufficient dynamic information processing and edge detail capture in colorectal polyp image segmentation, such as boundary information loss and wrong segmentation, this paper proposes a colorectal polyp segmentation method based on Swin Transformer framework. Firstly, Transformer encoder is used to extract the pathological features of the image step by step. Secondly, the improved second-order channel attention (SOCA) mechanism is used to enhance cross-level information interaction ability and effectively extract rich multi-scale context feature information. Furthermore, the discrete cosine transform (DCT) in the attention mechanism of reverse frequency channel is used to highlight the channel characteristics of multi-scale context information. Finally, the image features are enhanced from both dynamic and static depth through the cross-cue fusion (CCF) module to improve the dynamic information processing and detail capture capabilities. When tested on the datasets CVC-ClinicDB, Kvasir, CVC-ColonDB, and ETIS-LaribPolypDB, Dice indices are 0.942, 0.924, 0.800 and 0.774, respectively. The MIoU indices are 0.896, 0.878, 0.726 and 0.697, respectively. The experimental data show that the proposed method can effectively segment colorectal polyp images and provide a new idea for the diagnosis of colorectal polyp.

A new target detection method GhostNet YOLO v7 (G-YOLO v7) based on GhostNet and attention mechanism with large detection head network structure is proposed to solve the problems of high missed detection rate, low detection success rate and large model volume of traditional unmanned aerial vehicle (UAV) target detection algorithm. This technology adds a large 160×160 target detection head on the basis of YOLO v7-tiny to improve the small target detection ability, and lightweight processing is performed on the network. The original 20×20 minimum detection head and its convolution structure are deleted, and GhostNet convolution module is added to reduce the number of network parameters and model volume. At the same time, the loss function is modified to wise intersection over union (WIoU), and parallel convolutional block attention module (PCBAM) is added to improve the detection accuracy. The experimental results show that the mAP@0.5 of target detection based on G-YOLO v7 network structure is 42.3%, which is 5.2% higher than that of YOLO v7-tiny, 7.4% higher than that of YOLO v8n. The parameter quantity and model volume of G-YOLO v7 are only 33.9% and 37.9% of YOLO v7-tiny respectively, 64% and 75.6% of YOLO v8n respectively, which can be effectively applied to unmanned aerial vehicle aerial image target detection.

In response to issues such as blurriness, low brightness, and poor contrast in underwater bridge pier crack images, this paper proposes an image enhancement method based on image fusion. Firstly, homomorphic filtering is applied to denoise the crack images, eliminating the noise impact of the underwater environment. Subsequently, the homomorphic filtering processed crack images are transformed from the red-green-blue (RGB) space to the lab color space (LAB). The L channel in the LAB is subjected to adaptive Gamma correction for brightness adjustment, and the processed crack images are then converted back to the RGB space, resulting in enhanced crack images (Image 1). The crack images processed by homomorphic filtering are also subjected to contrast enhancement through contrast limited adaptive histgram equalization (CLAHE), producing contrast-enhanced crack images (Image 2). Finally, Image 1 and Image 2 are weighted fused to obtain the ultimate enhanced image. Subjective visual effects and three objective evaluation metrics are employed to validate the reliability of the proposed analysis method. The results indicate that the proposed method effectively enhances the brightness, clarity, and contrast of the crack images processed in this paper.

Aiming at the problems of image contrast reduction and serious color cast in turbid water, we constructed a dataset of underwater image for experimental turbid water, and proposed an image enhancement method based on improved Shallow-UWnet network. Firstly, we employed the algorithm of gray scale for global color correction to original images. And then we utilized the improved Shallow-UWnet network, which learned the mapping relationship between the distorted and the normal images, to achieve underwater image enhancement. Finally, we improved the contrast of images to obtain final results, by employing contrast limited adaptive histogram equalization (CLAHE). The experimental results show that our method is superior to other 5 ones not only in subjective and objective evaluation indexes but also in key points matching. And it is effectively in correcting the color cast in different turbid water and improving the contrast and clarity. This method can be applied to underwater in-situ environment with turbidity, and is an available solution for improving underwater visualization. It has wide prospect in underwater detection, underwater salvation, underwater exploration and so on.

Aiming at the problem of insufficient feature extraction in bridge disease images under complex environmental background and noise, the method of integrating fractal geometric features with YOLOv7 network is proposed to improve the accuracy of disease detection. Firstly, the fractal feature module (FFM) is designed to obtain the fractal feature map of bridge disease images. Secondly, the adaptive feature fusion layer is designed to integrate the extracted fractal features into the YOLOv7 network and the network can obtain more expressive feature map. Finally, the coordinate attention mechanism is introduced to enhance the detection accuracy of the network for small diseases. The experiment examines the complex images of five bridge diseases including efflorescence, crack, exposedbars, corrosionstain and spallation. The results show that, with the same dataset and numbers of iteration, the mean average precision of YOLOv7 network increases from 82.94% to 86.24%, and the average accuracy of crack disease detection increases the most significantly, from 75.92% to 81.29%.

Aiming at the deficiency of the existing deep learning methods in the unified expression of global and local features of breast cancer histopathology images, a breast cancer histopathology image classification method based on multi-level deep feature fusion is proposed by combining the long-distance modeling Transformer and strong local perception convolutional neural network (CNN). This method uses the dual-branch parallel Deit-B and ResNet-18 model as the backbone architecture, and introduces the feature fusion operations in the middle layer and end position of the dual-branch network respectively, which effectively strengthens the joint learning of global and local deep features of breast cancer histopathology images. In addition, dense connection operations are introduced between the residual modules of CNN tributaries to improve the information transmission of intermediate layer fusion features. Through global-local feature extraction and feature interaction between and within tributaries, discriminative features for breast cancer histopathology image classification can be captured more effectively. The ablation experiment and comparative experimental results on the public dataset of breast cancer histopathology images BreakHis prove the effectiveness of the proposed method, and the optimal classification accuracy of 99.83% can be obtained.

Based on Thomson scattering theory and computer numerical simulation, the influence of circularly polarized laser pulse intensity variation on electron radiation energy spectrum, radiation characteristics in peak radiation direction and space spectrum is studied in this paper. The results show that the spectrum of both backward radiation and peak radiation is shifted and broadened, and this phenomenon becomes more significant with the increase of laser intensity. There is only fundamental wave in the backward radiation spectrum. With the increase of laser intensity, the peak amplitude decreases, and the radiation value is gradually continuous. The peak radiation will produce high order harmonics. The peak amplitude is positively correlated with the laser intensity. As the frequency shift increases, the spectrum overlaps. What′s more, the frequency shift results in continuous spectrum under the action of super strong laser. We further discuss the electron radiation characteristics in the direction of the peak radiation power. We find that the pulse width of the radiation power and the polar angle of the maximum radiation direction have similar characteristics with the change of laser intensity, and both jump at some same special laser intensity. The spatial distribution of electron radiation power is symmetrical in the direction of propagation, and this symmetry is broken when the laser intensity is too large. The results of this study provide an opportunity for further research on the influence of circularly polarized laser on electron radiation characteristics.

In underwater wireless optical communication systems, the effects of water absorption, scattering, and turbulence make channel estimation and signal detection different, leading to increased communication bit error rates (BER) and even communication failure. To address the difficulties of channel estimation and signal detection in complex underwater channels for optical communication, a machine learning (ML)-based channel estimation and demodulation algorithm is proposed, and its performance in underwater channel estimation and signal detection in direct current biased optical-orthogonal frequency division multiplexing (DCO-OFDM)optical communication systems is studied. Firstly, based on the proposed channel estimation and demodulation algorithm (deep neural network (DNN) and unsupervised learning k-means constellation demodulator), simulation modeling of complex channel frequency response, second-order equalization, and bit error analysis are completed. Secondly, studies on the signal-to-noise ratio (SNR) gains in complex optical communication channels are conducted, comparing traditional least squares (LS), linear minimum mean square error (LMMSE) channel estimation algorithms, and minimum distance demodulation algorithms. In the simulation results, in an underwater channel with a turbulence scintillation index of 0.18 and a distance of 10 m, the proposed channel estimation algorithm provides a signal-to-noise ratio gain larger than 6 dB and 1 dB compared with the LS and LMMSE estimation for 8-order quadrature amplitude modulation (8-QAM) subcarriers at a bit error rate of 10-5. Additionally, using the proposed signal detection algorithm, an SNR gain larger than 1 dB is achieved compared with traditional algorithms. The simulation results demonstrate that the proposed ML-based channel estimation and demodulation algorithm can impraove the performance of complex underwater optical communication channels. The research results provide a reference for the design of long-distance, high-speed complex underwater optical communication systems.

In this paper, a nanosecond pulsed laser was used for laser cleaning of acrylic urethane paint on the surface of 7050 aluminum alloy, and the effects of laser power, scanning speed and repetition frequency on the paint removal rate and surface roughness were investigated. Quantitative analysis of the paint removal rate was achieved by binarizing the super depth of field image of the substrate surface. The results show that as the laser power increases, the paint removal rate gradually increases and the surface roughness first decreases and then increases. As the scanning speed and repetition frequency increase, the paint removal rate increases and then decreases, and the surface roughness decreases and then increases. A generalized regression neural network (GRNN) model was used to establish the correlation density function between laser process parameters and cleaning quality. The best combination of parameters for the laser paint removal process was obtained by multi-objective optimization of the model through the multi-objective sparrow search algorithm (MOSSA). With this laser process parameter, the paint removal rate was 99.16% and the surface roughness was 1.32 m.

A polyp segmentation method based on dual path feature multi-scale subtraction is proposed to address the issues of significant size differences, unclear boundaries, and scattered distribution of colon polyps. The main branch merges adjacent feature maps by reconstruction subtraction units and attention models, enhancing the boundary information of polyps and the ability to extract polyp features. Simultaneously, a learnable visual center (LVC) is introduced to aggregate local corner key regions of the input image. In the sub-branch, a multi-scale extraction module and a conv-transpose sample module are designed to fuse into an aggregation module (AGG) for multi-scale size polyp extraction, restoring and supplementing more detailed information. The proposed method is experimentally analyzed on four public datasets, and the experimental results show that our method has good generalization performance on polyp segmentation, with mDice and mIoU achieving 93.28% and 88.98% on the CVC-ClinicDB dataset, respectively.