Journals >Optics and Precision Engineering
Contents
2024
Volume: 32 Issue 9
15 Article(s)
Export citation format
[in Chinese]
High thrust focused magnetic linear motor for needling massive transfer platform
Qiang LIU, Ning MA, Wei WANG, Bangcheng HAN, and Pingjuan NIU
A high thrust concentrated magnetic linear motor is proposed to address the problem of low thrust and obvious edge effects leading to slow transfer speed in needle punched massive transfer platforms. The outer V-shaped structure is used to suppress leakage magnetic effect， and the double-layer Halbach structure is usedA high thrust concentrated magnetic linear motor is proposed to address the problem of low thrust and obvious edge effects leading to slow transfer speed in needle punched massive transfer platforms. The outer V-shaped structure is used to suppress leakage magnetic effect， and the double-layer Halbach structure is used to improve air gap magnetic density. The finite element method is used to compare and analyse the air gap magnetic density and leakage magnetic flux of the three schemes. The air gap magnetic flux and leakage magnetic flux of the V-shaped double-layer Halbach scheme are 0.6 T and 0.05 T， respectively. Compared with the single-layer trapezoidal scheme， the air gap magnetic flux is increased by 17.6%， and the leakage magnetic flux is reduced by 71.4%. The equivalent magnetization current method was used to segment the magnetic field of the motor， and a rectangular boundary condition was established using the segmentation compensation method. A mathematical model of the magnetic circuit of the linear motor was established， and the key structural parameters affecting the linearity of the motor thrust were obtained. The shape and structural parameters were optimized based on the finite element method. The results show that after optimization， the maximum and minimum magnetic densities of the air gap are 0.67 T and 0.47 T， respectively. The change rates of magnetic densities in the x and y directions are 25% and 29.8%， respectively. The thrust simulation calculation result is 278 N， the thrust fluctuation rate is 2.1%， the no-load acceleration is 10 m/s2， and the theoretical chip transfer speed can reach 50 pieces/s. Based on the optimization results， a linear motor was developed for the massive transfer platform. The test results are consistent with the optimization results， which is of great significance for improving the transfer efficiency of the massive transfer platform..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1347 (2024)
Mechanism for laser-induced damage bad chip of Micro-LED and optimization of processing parameters
Jian QIAO, Zhenduo WU, Xinhan PENG, Yuxuan RAN, and Jingwei YANG
To improve the manufacturing quality of Micro-LED display panels, it is crucial to perform in-situ removal and repair of defective chips. This study introduces a laser two-temperature ablation model developed with COMSOL to explore the mechanisms and process parameters for eliminating defective chips using a laser. IniTo improve the manufacturing quality of Micro-LED display panels, it is crucial to perform in-situ removal and repair of defective chips. This study introduces a laser two-temperature ablation model developed with COMSOL to explore the mechanisms and process parameters for eliminating defective chips using a laser. Initially, the ablation threshold of the Micro-LED was identified by measuring the square diameter of the surface spot created by a single-pulse laser ablation. A fitting analysis was then carried out to determine the relationship between laser parameters and ablation depth, confirming the accuracy of the two-temperature ablation model. An extensive analysis was also conducted to examine the laser spot overlap rate, laser fluence, ablation characteristics, and removal mechanisms for different scanning paths on the affected chips. Utilizing the two-temperature ablation model, the laser scanning path and process parameters were optimized. The findings reveal that the femtosecond laser ablation model for Micro-LED has a maximum analytical error of 9.28%. The optimized laser removal method allows for rapid and precise removal at a rate of 1 s/chip, while ensuring high-quality repair of pad surfaces. This provides essential guidance for efficient repair processes in large-scale production environments of large-format Micro⁃LED display panels..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1360 (2024)
Micro-optical illumination properties of hierarchical microlens arrays formed by hot-embossing
Manfeng HU, Lei LI, and Jin XIE
In LED lighting, glare from uneven light distribution, along with ultraviolet and blue light, can lead to visual fatigue and eye damage. Currently, adjustments are primarily made to the light source, and modifying the surface structure at the nanoscale is costly. To address this, a microscope-scale hierarchical microleIn LED lighting, glare from uneven light distribution, along with ultraviolet and blue light, can lead to visual fatigue and eye damage. Currently, adjustments are primarily made to the light source, and modifying the surface structure at the nanoscale is costly. To address this, a microscope-scale hierarchical microlens array light guide plate (LGP) has been developed using hot-embossing rapid prototyping technology to examine the impact of the scale of hierarchical microlenses on health lighting. The study first analyzes the relationship between the microlens aspect ratio and the distribution of the reflected electric field on the surface of the microprism. It then explores how the microstructure parameters of the LGP affect diffuse and specular reflection, as well as transmission. The study further compares the effects of the microscale of hierarchical microlenses on micro-optical properties such as illuminance and uniformity with existing market solutions like screen-printed microdots and microprism arrays. Findings indicate that the reflected light monitoring surface for ultraviolet and blue light shows a smaller variation in the reflection electric field when the microlens aspect ratio is between 0.6-1, suggesting higher diffuse reflection. Experiments demonstrate that the hierarchical microlens array LGP, with microlens heights of 5-30 μm and microprism heights of 10-80 μm, can be produced in just 4 seconds. Illuminance uniformity improves as the microlens aspect ratio increases, with illuminance 45% and 83% higher than that of the microprism and screen-printed microdot, respectively. Moreover, compared to the microprism, the diffuse reflectance of ultraviolet and blue light is 8% and 18% higher, respectively, and the mirror reflectance is 80%⁃83% lower. Thus, controlling the aspect ratio of the hierarchical microlens can effectively reduce lighting glare and mitigate the effects of ultraviolet and blue light..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1371 (2024)
Fabrication of glass microprism via interfacial erosion induced quasi-anisotropic wet etching
Feier LI, Jiajia YU, Liqun DU, Mengxi WU, and Junshan LIU
Microprism arrays are extensively utilized across various industries. Glass-based microprisms, in comparison to polymers, offer superior corrosion resistance, thermal stability, and durability. However, the isotropic nature of glass makes micromachining into prism-like structures challenging. In this paper, we introducMicroprism arrays are extensively utilized across various industries. Glass-based microprisms, in comparison to polymers, offer superior corrosion resistance, thermal stability, and durability. However, the isotropic nature of glass makes micromachining into prism-like structures challenging. In this paper, we introduce a novel interfacial erosion-induced quasi-anisotropic wet etching technique to manufacture glass microprism arrays efficiently and cost-effectively. For the first time, interfacial erosion is analyzed within a metacellular automaton to study the wet etching process, uncovering the dynamics between side etching and interfacial erosion and their impact on the side wall profiles. This approach yields varied microstructures with tilted morphologies. Leveraging quasi-anisotropic etching characteristics, we successfully produce microprism arrays with adjustable spacing, shape, and size, achieving 98% repeatability between arrays. Our method significantly enhances LED lamp diffusion, increasing average brightness by 4.6 times. This research not only pioneers a new direction in glass wet etching but also incorporates quasi-anisotropic properties into the traditional isotropic framework, presenting a straightforward, economical technique for fabricating glass microprism arrays and similar devices..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1384 (2024)
[in Chinese]
Large field optical imaging system based on freeform surface optics
Yupeng XIONG, Shiyu CHEN, Cheng HUANG, Wenwen LU, Xiaoqiang PENG, Shanyong CHEN, Tao LAI, and Yang OU
To achieve extensive field of view optical imaging, research has been conducted on a system utilizing free-form surface optical components. This research encompasses the optical design of such systems, ultra-precision machining of free-form surfaces, synchronous error detection in form and position, and system integratTo achieve extensive field of view optical imaging, research has been conducted on a system utilizing free-form surface optical components. This research encompasses the optical design of such systems, ultra-precision machining of free-form surfaces, synchronous error detection in form and position, and system integration with imaging experiments. Initially, by studying the mechanism of aberration formation under multiple constraints, a precise model for the parametric description of free-form surface optical systems was developed. Subsequently, nano-precision machining and high-frequency suppression techniques were explored for aluminum alloy free-form mirrors, alongside high-precision shape and position detection using computer-generated holographic elements. Comprehensive experiments on system integration and imaging were performed. The results show that the system's field of view is 30°×5°, with an optical transfer function value exceeding 0.7, approaching the diffraction limit, and a maximum pixel root mean square radius of 2.075 μm. The surface shape accuracy of the free-form optical components has an RMS value better than 20 nm, and position accuracy better than 1 μm. Post-assembly, the system fulfills the criteria for large field of view and high-resolution applications, also demonstrating stability, reliability, and rapid response in manufacturing..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1261 (2024)
Design of fiber scanning Lissajous trajectory near-eye display system
Tong WU, Xiangyu YANG, Jincong ZOU, Youwen LIU, Ling WANG, Jiming WANG, Yuangang LU, and Chongjun HE
To further reduce the size of head-worn displays (HWDs), a near-eye display system utilizing an asymmetric fiber cantilever has been developed, alongside a specific pixel division rule. This study explores the factors affecting the scanning trajectory, image reconstruction capabilities, display resolution, and frame raTo further reduce the size of head-worn displays (HWDs), a near-eye display system utilizing an asymmetric fiber cantilever has been developed, alongside a specific pixel division rule. This study explores the factors affecting the scanning trajectory, image reconstruction capabilities, display resolution, and frame rate of the system. The resonant frequencies in orthogonal directions are tailored for enhanced definition and faster frame rates, with calculated structural parameters for the asymmetric fiber cantilever. The display unit measures 2 mm in diameter and 19 mm in length, offering a resolution of 160×120 pixels at a 10 Hz frame rate, with an image reconstruction error rate of 0.002%. It can potentially reach a maximum resolution of 290×220 pixels and a frame rate of 95 Hz. A fiber-scanning near-eye display system has been constructed, employing a camera to document the experimental outcomes. These results confirm the display's efficacy and the promising future applications of asymmetric fiber cantilevers in HWDs..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1273 (2024)
Temperature insensitive optical fiber dual-cavity salinity sensor
Yitong LI, Yuqiang YANG, Xiaoguang MU, Ji WANG, and Jiale GAO
Addressing the issue of temperature cross-sensitivity in fiber optic microcavity salinity sensors, a novel design featuring an open saltwater cavity constructed from a ceramic ferrule and zirconia sleeve has been developed. The positive thermal expansion coefficient of zirconia effectively counteracts the negative therAddressing the issue of temperature cross-sensitivity in fiber optic microcavity salinity sensors, a novel design featuring an open saltwater cavity constructed from a ceramic ferrule and zirconia sleeve has been developed. The positive thermal expansion coefficient of zirconia effectively counteracts the negative thermal and optical coefficients of seawater, thereby compensating for temperature fluctuations in salinity measurements. To enhance salinity detection sensitivity, the open saltwater cavity is paired in parallel with a closed air cavity. This configuration precisely controls the free spectral range of both cavities, inducing a vernier effect that significantly boosts salinity sensitivity through this phenomenon. Experimental data indicate that the temperature sensitivity of the newly designed saltwater cavity is just 0.025 nm/℃, substantially lower than the -0.35 nm/℃ sensitivity due to seawater's thermal and optical properties. Furthermore, within a salinity range of 0 to 3%, the sensitivity of the dual cavity system reaches 0.180 8 nm/%, which is 10.5 times greater than that of a single saltwater cavity. The sensor offers benefits such as ease of manufacturing, high sensitivity, and robust stability. Complete temperature compensation is theoretically achievable with this setup..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1283 (2024)
Research developments of extreme ultra-violet multilayers for 40-90 nm
Bo LAI, Li JIANG, Runze Qi, and Zhanshan WANG
The multilayer mirrors in the extreme ultraviolet region have significantly advanced various scientific and technological fields, including extreme ultraviolet lithography, astrophysics, plasma diagnosis, attophysics, and free electron lasers. These multilayer coatings are crucial for developing mirrors that achieve hiThe multilayer mirrors in the extreme ultraviolet region have significantly advanced various scientific and technological fields, including extreme ultraviolet lithography, astrophysics, plasma diagnosis, attophysics, and free electron lasers. These multilayer coatings are crucial for developing mirrors that achieve high reflectivity and narrow bandwidth at normal incidence within the extreme ultraviolet spectrum. Multilayer mirrors are now key components in both imaging and spectroscopy across various applications. Optical imaging systems equipped with multilayer mirrors have conducted numerous solar observations in different emission lines, achieving unprecedented spatial and spectral resolutions. This paper reviews the progress of the most promising multilayer mirrors, covering the 40 nm to 90 nm range. Following a brief introduction to multilayer theory, it highlights the most promising material pairs and layer stack structures based on Mg, Sc, and lanthanides. The review emphasizes multilayer mirrors that exhibit high reflectivity within the 40 nm to 90 nm wavelength range and discusses their temporal stability..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1293 (2024)
Spatiotemporal distribution characteristics analysis of local flameout in jet flames based on 5 kHz planar laser-induced fluorescence
Long GAO, Jiangbo PENG, Xin YU, Chaobo YANG, Zhen CAO, Xun YUAN, and Wenbei LIU
Studying the distribution characteristics of local flameout using high-speed optical measurement methods is crucial for understanding jet flame instability mechanisms. Initially, a planar laser-induced fluorescence (PLIF) measurement system operating at 5 kHz is set up. An experimental study on high-speed jet flames isStudying the distribution characteristics of local flameout using high-speed optical measurement methods is crucial for understanding jet flame instability mechanisms. Initially, a planar laser-induced fluorescence (PLIF) measurement system operating at 5 kHz is set up. An experimental study on high-speed jet flames is conducted with OH-PLIF to monitor the dynamic development of the flame structure. Image processing techniques like bilateral filtering and local adaptive thresholding with vertical sliding windows are then applied to preprocess PLIF images. Innovative features such as the length and spatial location of local fracture structures are extracted. Subsequently, the temporal and spatial distribution characteristics of local flameouts are analyzed, enhancing the high-speed PLIF diagnostic and analytical methods. Results indicate that the bilateral filtering algorithm significantly improves image quality, achieving a peak signal-to-noise ratio of 34.3 dB and a structural similarity index of 0.93. The algorithm effectively reduces noise levels. The F1 score for the adaptive image segmentation method using vertical sliding windows reaches 93.30%. Moreover, the count of fractured structures linearly increases over time. As the jet Mach number rises from 0.5 to 1.6, the number of fractured structures per second in the unilateral image jumps from 790.7 to 9 217.2, showing an exponential relationship with the jet Mach number. Predominant areas for fracture structures include the flame arms and the upper central fuel zone. This research validates the effectiveness of high-speed PLIF technology in assessing the spatial and temporal distribution of local flameout, broadening its potential applications..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1307 (2024)
Laser confocal interference surface rapid and high-precision automatic measurement
Jiahao ZHAO, Shuai YANG, Zheng YANG, Jian CUI, Lirong QIU, and Weiqian ZHAO
Considering the low accuracy and slow speed of measuring objects with spherical surfaces using traditional interferometry due to manual adjustments, an automatic laser confocal interferometric method is proposed for rapid and high-precision measurement of surface shapes. This method utilizes precise confocal fixed-focuConsidering the low accuracy and slow speed of measuring objects with spherical surfaces using traditional interferometry due to manual adjustments, an automatic laser confocal interferometric method is proposed for rapid and high-precision measurement of surface shapes. This method utilizes precise confocal fixed-focus technology for accurate confocal axial positioning of the object. The object's posture is adjusted using a five-dimensional workbench. Zernike fitting is employed to minimize defocus and tilt errors, achieving zero-order fringe adjustment and automatic determination of the surface shape. A laser confocal interferometric measurement system was developed for this purpose. Experimental results show that the difference between the automatic interference measurement results and those of the Zygo interferometer is 0.000 6λ, with a repeatability accuracy of 0.001 9λ. The average measurement time for a single surface is 16.5 s, improving measurement efficiency by 1.56 times compared to manual adjustments. This method offers an effective solution for the rapid and precise measurement of interferometric surface shapes..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1320 (2024)
Mid-infrared laser absorption spectroscopy CO2 spectrometer
Jingyi LIU, Yue PAN, Lu YAO, Liting YU, Hui LI, and Ruifeng KAN
Carbon Dioxide (CO2), as the most crucial greenhouse gas, significantly influences the regulation of Earth's climate and ecosystems. Accurately detecting trace CO2 concentrations is vital for assessing the greenhouse effect. Gas molecules exhibit strong absorption in the mid-infrared range, which is advantageous foCarbon Dioxide (CO2), as the most crucial greenhouse gas, significantly influences the regulation of Earth's climate and ecosystems. Accurately detecting trace CO2 concentrations is vital for assessing the greenhouse effect. Gas molecules exhibit strong absorption in the mid-infrared range, which is advantageous for highly sensitive trace gas detection. This paper describes the design of a compact CO2 monitoring instrument using an interband cascade laser (ICL) at 4.219 μm as the light source, paired with Tunable Diode Laser Absorption Spectroscopy (TDLAS). The instrument was calibrated with various standard gas concentrations and, using Allan variance analysis, achieved a detection limit of 16.6×10-9 with an integration time of 8.7 s, meeting the requirements for sensitive CO2 measurements in atmospheric monitoring. Additionally, the instrument was employed to monitor atmospheric CO2 over eight consecutive days and was benchmarked against the commercial LGR instrument, confirming the reliability of our miniaturized CO2 monitor and the effectiveness of combining mid-infrared light source with TDLAS technology for precise atmospheric CO2 measurement. The findings offer a technical reference for high-precision CO2 concentration detection, while the system's simple structure, compact size, low power consumption, and lightweight design are ideal for developing CO2 monitoring capabilities on airborne platforms..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1330 (2024)
System parameter calibration based on a specular and diffuse calibration plate
Xiaona YANG, Yubo NI, Zhaozong MENG, Nan GAO, Zeqing YANG, Guofeng ZHANG, Wei YIN, Hongwei ZHAO, and Zonghua ZHANG
To achieve full-field measurement of composite surface objects and expand the measurement range, this paper proposes a system parameter calibration method using a specular and diffuse calibration plate. This method captures phase information and calibration data for both specular and diffuse reflections at the same pixTo achieve full-field measurement of composite surface objects and expand the measurement range, this paper proposes a system parameter calibration method using a specular and diffuse calibration plate. This method captures phase information and calibration data for both specular and diffuse reflections at the same pixel point. Initially, a specular and diffuse calibration plate with a checkerboard pattern and known marker distances is designed and manufactured. Using the unique configuration of the calibration plate, the complete reference phase is derived through external parameter constraints and spline interpolation, facilitated by limit translation. Subsequently, spatial distance, system parameters, and the relationship between pixel position and horizontal coordinates XY are calibrated pixel-by-pixel within the same coordinate system. The 3D morphology of the composite surface object is then reconstructed using the complete reference phase and pixel-by-pixel calibration data. Experimental results indicate that this method extends the measurement range by 1.5-2 times compared to traditional non-full-field methods. The proposed approach not only maintains measurement accuracy but also enables full-field measurement of 3D morphology of discontinuous composite surfaces, proving to be of significant practical value for measuring the 3D morphology of composite surface objects..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1338 (2024)
[in Chinese]
Active learning-clustering-group convolutions network for hyperspectral images classification
Jing LIU, Yinqiao LI, and Yi LIU
Hyperspectral image (HSI) classification using convolutional neural networks often grapples with a large number of network parameters and a scarcity of class-labeled samples. To tackle these issues, we propose a method called AL-CGNet, which integrates active learning and clustering with group convolutions network for Hyperspectral image (HSI) classification using convolutional neural networks often grapples with a large number of network parameters and a scarcity of class-labeled samples. To tackle these issues, we propose a method called AL-CGNet, which integrates active learning and clustering with group convolutions network for efficient HSI classification. AL-CGNet combines a convolutional neural network with active learning and clustering to enhance feature extraction and classification, while a group convolutions-based lightweight network model significantly reduces parameter count. Initially, HSI reduced in dimensionality through linear discriminant analysis is segmented into clusters via the mini-batch K-means algorithm. The central feature of each cluster substitutes the samples within, leveraging information from unlabeled samples. Subsequently, feature maps are segmented into groups along the spectral dimension in the group convolutions network, where each group sequentially extracts spatial-spectral features through multiple residual blocks. This grouping strategy optimizes band redundancy and diversity, cuts down network parameters, and achieves lightweighting. Active learning then selects informative samples for the training set, mitigating the issue of limited labeled samples. Experimental results demonstrate that AL-CGNet, with only 6% training samples, significantly outperforms ClusterCNN, SSRN, and HybridSN on the Indian Pines, Botswana, and Houston datasets, achieving overall accuracies of 99.57%, 99.23%, and 98.82%, respectively. Remarkably, AL-CGNet remains effective even with a smaller training sample size of 5%. This method not only boosts HSI classification efficiency but also ensures robust feature extraction and high accuracy..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1395 (2024)
Regional gradient dimming strategy for low-power silicon-based OLED microdisplays
Yuan JI, Xinde MA, Yanrui SUN, Baoliang CHEN, and Yin ZHANG
In addressing the high power consumption of near-eye display devices, this paper introduces a region gradient dimming algorithm that adapts to the human eye's gaze point, informed by studies on human visual characteristics. Initially, a central fovea-brightness masking experiment established the Just Noticeable DifIn addressing the high power consumption of near-eye display devices, this paper introduces a region gradient dimming algorithm that adapts to the human eye's gaze point, informed by studies on human visual characteristics. Initially, a central fovea-brightness masking experiment established the Just Noticeable Difference (JND) threshold for the human eye. This data refined the traditional geometric optical path model of the eye gaze point, resulting in a model that aligns more closely with the eye's adaptive characteristics. The method significantly improved the efficiency of viewing angle calculations through a maximum viewing angle discrimination approach. An enhanced contrast algorithm was then applied during image preprocessing to improve the display quality while preserving the image's average brightness. Utilizing the JND threshold and gaze point data, the algorithm applied both regional and global power consumption limits to the image, reducing power usage while maintaining subjective visual perception. The algorithm underwent validation on an FPGA hardware platform, demonstrating a reduction in display power consumption for silicon-based Organic Light Emitting Diode (OLED) microdisplays by up to 23.05% on the Kodak standard test set. This achievement suggests that the low-power requirements for silicon-based OLED microdisplays are achievable, offering insights for performance enhancements..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1408 (2024)
Segmentation network for metastatic lymph nodes of head and neck tumors
Tao ZHOU, Daozong SHI, Jiawen XUE, Caiyue PENG, Pei DANG, and Zhongwei ZHOU
Head and neck tumors are prevalent malignant tumors in China, with prognosis significantly influenced by cervical lymph node metastasis. In medical practice, magnetic resonance imaging (MRI) is employed to identify metastatic lymph nodes. However, MRI images often suffer from blurred edges and low contrast between the Head and neck tumors are prevalent malignant tumors in China, with prognosis significantly influenced by cervical lymph node metastasis. In medical practice, magnetic resonance imaging (MRI) is employed to identify metastatic lymph nodes. However, MRI images often suffer from blurred edges and low contrast between the lesion and surrounding tissue. This paper introduces a segmentation network tailored for metastatic lymph nodes in head and neck tumors. Initially, a cross-layer and cross-field attention module is developed, integrating features from both deep and shallow layers to enhance the shape representation of metastatic lymph nodes through a self-attention mechanism. This module improves contextual semantic understanding across different receptive fields, allowing for pixel-level fusion of shallow and deep feature maps, thereby enhancing the morphological details of metastatic lymphatic nodes. Subsequently, a multi-scale feature fusion module is designed to amalgamate features across various scales in the feature pyramid, enriching the morphological details of the lymph nodes. Furthermore, an enhanced attention prediction head module is implemented, combining parallel self-attention and gate channel transformation to accentuate the lesion area and refine its boundaries on the feature map. The network's effectiveness is confirmed using a clinical dataset of lymph node metastasis medical images. The performance metrics, APdet, APseg, ARdet, ARseg, mAPdet, and mAPseg for lymph node metastasis lesion segmentation are 74.88%, 74.12%, 63.11%, 62.28%, 74.64%, and 74.04%, respectively. This network provides precise detection and segmentation of lymph node metastasis lesions, offering significant benefits for lymph node diagnosis..
Optics and Precision Engineering
- Publication Date: May. 10, 2024
- Vol. 32, Issue 9, 1420 (2024)
© Copyright 2018-2021 | Chinese Laser Press. All Rights Reserved 沪ICP备15018463号-20