Multi-class classification of pathological myopia based on fundus photography

Pathological myopia(PM)is a severe ocular disease leading to blindness.As a traditional noninvasive diagnostic method,fundus color photography(FCP)is widely used in detecting PM due to its highfidelity and precision.However,manual examination of fundus photographs for PM is time-consuming and prone to high error rates.Existing automated detection technologies have yet to study the detailed classification in diagnosing different stages of PM lesions.In this paper,we proposed an intelligent system which utilized Resnet101 technology to multi-categorically diagnose PM by classifying FCPs with different stages of lesions.The system subdivided different stages of PM into eight subcategories,aiming to enhance the precision and efficiency of the diagnostic process.It achieved an average accuracy rate of 98.86%in detection of PM,with an area under the curve(AUC)of 98.96%.For the eight subcategories of PM,the detection accuracy reached 99.63%,with an AUC of 99.98%.Compared with other widely used multi-class models such as VGG16,Vision Transformer(VIT),EfficientNet,this system demonstrates higher accuracy and AUC.This artificial intelligence system is designed to be easily integrated into existing clinical diagnostic tools,providing an efficient solution for large-scale PM screening.

Warhead fragments motion trajectories tracking and spatio-temporal distribution reconstruction method based on high-speed stereo photography

High speed photography technique is potentially the most effective way to measure the motion parameter of warhead fragment benefiting from its advantages of high accuracy,high resolution and high efficiency.However,it faces challenge in dense objects tracking and 3D trajectories reconstruction due to the characteristics of small size and dense distribution of fragment swarm.To address these challenges,this work presents a warhead fragments motion trajectories tracking and spatio-temporal distribution reconstruction method based on high-speed stereo photography.Firstly,background difference algorithm is utilized to extract the center and area of each fragment in the image sequence.Subsequently,a multi-object tracking(MOT)algorithm using Kalman filtering and Hungarian optimal assignment is developed to realize real-time and robust trajectories tracking of fragment swarm.To reconstruct 3D motion trajectories,a global stereo trajectories matching strategy is presented,which takes advantages of epipolar constraint and continuity constraint to correctly retrieve stereo correspondence followed by 3D trajectories refinement using polynomial fitting.Finally,the simulation and experimental results demonstrate that the proposed method can accurately track the motion trajectories and reconstruct the spatio-temporal distribution of 1.0×10^(3)fragments in a field of view(FOV)of 3.2 m×2.5 m,and the accuracy of the velocity estimation can achieve 98.6%.

A Systematic Literature Review of Machine Learning and Deep Learning Approaches for Spectral Image Classification in Agricultural Applications Using Aerial Photography

Recently,there has been a notable surge of interest in scientific research regarding spectral images.The potential of these images to revolutionize the digital photography industry,like aerial photography through Unmanned Aerial Vehicles(UAVs),has captured considerable attention.One encouraging aspect is their combination with machine learning and deep learning algorithms,which have demonstrated remarkable outcomes in image classification.As a result of this powerful amalgamation,the adoption of spectral images has experienced exponential growth across various domains,with agriculture being one of the prominent beneficiaries.This paper presents an extensive survey encompassing multispectral and hyperspectral images,focusing on their applications for classification challenges in diverse agricultural areas,including plants,grains,fruits,and vegetables.By meticulously examining primary studies,we delve into the specific agricultural domains where multispectral and hyperspectral images have found practical use.Additionally,our attention is directed towards utilizing machine learning techniques for effectively classifying hyperspectral images within the agricultural context.The findings of our investigation reveal that deep learning and support vector machines have emerged as widely employed methods for hyperspectral image classification in agriculture.Nevertheless,we also shed light on the various issues and limitations of working with spectral images.This comprehensive analysis aims to provide valuable insights into the current state of spectral imaging in agriculture and its potential for future advancements.

耦合超声场的气液流动特性

功率超声作为一种新兴的绿色技术,越来越受到石油行业的重视。结合稠油超声降黏机制,从气泡瞬态演化规律和溃灭特性出发,对超声作用下空化气泡和空化流场进行数值模拟和实验验证。模拟结果表明,增加超声频率和减小气泡之间的距离会导致单个空化气泡最大半径的减小,气泡个数的增加会使气泡生长至破裂过程的速度减小,延长超声作用时间和增大超声频率可增加超声场中最大空化压力。实验表明,空化气泡不断在超声场中振荡并互相聚并,且随着超声作用时间的延长,该过程更加剧烈。研究内容为稠油超声降黏在石油化工行业的应用提供了重要的理论指导。

Fundus photography,fundus fluorescein angiography,and optical coherence tomography of healthy cynomolgus monkey,New Zealand rabbit,Sprague Dawley rat,and BALB/c mouse retinas

Background:A variety of experimental animal models are used in basic ophthalmological research to elucidate physiological mechanisms of vision and disease pathogenesis.The choice of animal model is based on the measurability of specific parameters or structures,the applicability of clinical measurement technologies,and the similarity to human eye function.Studies of eye pathology usually compare optical parameters between a healthy and altered state,so accurate baseline assessments are critical,but few reports have comprehensively examined the normal anatomical structures and physiological functions in these models.Methods:Three cynomolgus monkeys,six New Zealand rabbits,ten Sprague Dawley(SD)rats,and BALB/c mice were examined by fundus photography(FP),fundus fluorescein angiography(FFA),and optical coherence tomography(OCT).Results:Most retinal structures of cynomolgus monkey were anatomically similar to the corresponding human structures as revealed by FP,FFA,and OCT.New Zealand rabbits have large eyeballs,but they have large optic disc and myelinated retinal nerve fibers in their retinas,and the growth pattern of retinal vessels were also different to the human retinas.Unlike monkeys and rabbits,the retinal vessels of SD rats and BALB/c mice were widely distributed and clear.The OCT performance of them were similar with human beings except the macular.Conclusions:Monkey is a good model to study changes in retinal structure associated with fundus disease,rabbits are not suitable for studies on retinal vessel diseases and optic nerve diseases,and rats and mice are good models for retinal vascular diseases.These measures will help guide the choice of model and measurement technology and reduce the number of experimental animals required.

Fundus photography,fluorescein angiography,optical coherence tomography and electroretinography of preclinical animal models of ocular diseases

The eye is an immune-privileged and sensory organ in humans and animals.Anatomical,physiological,and pathobiological features share significant similarities across divergent species(1).Each compartment of the eye has a unique structure and function.The anterior and posterior compartments of the eye contain endothelium(cornea),epithelium(cornea,ciliary body,iris),muscle(ciliary body),vitreous and neuronal(retina)tissues,which make the eye suitable to evaluate efficacy and safety of tissue specific drugs(2).

基于无人机倾斜摄影测量技术与SVR算法的开采沉陷监测与预计

为实现矿区开采沉陷高精度监测与预计,将无人机倾斜摄影测量技术与支持向量回归算法(Support Vector Regression Algorithm,SVR)相结合,对赤峰市六家煤矿采区沉降量进行了监测与预计,并将预计结果与GPS实测结果进行了对比。结果表明:GPS实测矿区10 d测量沉降量为0.2~15.8 mm,东西南北区平均沉降量分别为51.2、31.5、46.6、130.8 mm,北侧沉降量分别为东、西、南侧的2.55、4.15、2.81倍,矿区北侧为沉降严重区,需要进行重点监测并制定合理有效的防治措施。基于无人机倾斜摄影测量技术与SVR算法预测的矿区东西南北区累计平均沉降量分别为49.4、31.5、45.8、134.6 mm,所有测点2种测量方式所得相对误差均在5%以内,反映出所提方法预计精度较高,为实现矿区开采沉陷高精度监测与预计提供了参考。

A New Chromatic Vision:The Early Impact of Color Photography on the Representation of Architecture

The role of color photography in the representation of architecture is a subject little investigated by architectural historiography.The link between the color values of architectural design and its visual transmission in the early phase of modernism was certainly problematic.Color photography had an undeniable impact on architectural color in practice:color photographs in books and periodicals published between the 1940s and 1960s clearly influenced the use of color in architectural design.

短视频生活化叙事的逻辑与引导

以手机为代表的智能终端作为生活化、具身化的媒介工具,催生一种新型的短视频叙事——生活化叙事。它源于人类对“身体在场”的不懈追求,具有强烈的“身体在场”特征,以期实现人与人的“技术性亲密”。它以人格化摄影与第一人称叙事、“自导自演”为手段的生活展演和镜头自在运动的视听语言风格为主要表征,制造出新的“身体在场”与令人惊诧的影像身体奇观。生活化叙事及其身体展演性是移动时代自我支配技术的新产物。在全民化叙事生产突破传统视听语法、贴近日常生活的同时,人们必须反思这种身体强势介入叙事中的逻辑失衡和过度产消问题,学会在新媒介生活中“自我照顾”和安生。

基于旋转等变卷积的航拍红外图像目标识别算法

为提高传统无人机红外目标识别算法对输入图像的旋转鲁棒性,提出一种具有旋转等变性的红外图像目标识别算法。参照可见光三通道结构,将红外图像扩张为三通道以丰富输入图像的细节及边缘信息;以旋转等变卷积为基础,设计并实现能够高度保留图像旋转特征的标准旋转等变卷积模块和旋转残差模块,使得所设计模型FC-YOLOv5对图像及图像中目标旋转具有鲁棒性;加入压缩和激励注意力机制自适应地学习到每个通道的重要性,并且根据任务的需要加权调整特征图中的通道贡献,提取重要的特征信息并抑制不重要的特征信息。在航拍行人车辆数据集和海上船舶数据集上验证模型的性能,以基准模型YOLOv5s及常见轻量级目标识别任务所用模型YOLOv8s、NanoDet作为对照组模型。实验结果表明,所提算法的平均精度均值相较于基准模型能够提升2%~4%,且当输入图像具有不同角度的旋转时,能够比对照组模型识别到更多旋转目标,且识别错误更少。

HkV和HRCT检查在职业性尘肺病诊断中的价值

目的探究高千伏胸片(HkV)和高分辨率计算机体层X线摄影术(HRCT)检查在职业性尘肺病诊断中的临床应用价值。方法回顾性分析133例进行职业病诊断的尘肺病患者资料,患者均接受HkV和HRCT检查,以临床病理检查为金标准,比较HkV和HRCT对尘肺患者肺部阴影的显示情况、图片质量、分期诊断符合率及尘肺合并症的检出情况。结果HRCT对肺部P影、小阴影检出率高于HkV(P<0.05),两组大阴影检出率比较,差异无统计学意义(P>0.05);HRCT检查图像质量有效率高于HkV检查(P<0.05);HRCT诊断总准确率高于HkV诊断(P<0.05);HRCT对肺部感染、肺气肿、胸膜增厚、气胸的检出率高于HkV(P<0.05),HRCT与HkV对肺结核、肺大泡检出率差异无统计学意义(P>0.05)。结论HRCT检查对职业性尘肺病患者肺部阴影、临床分期及合并症的诊断效果高于HkV,图片质量更高。

航空摄影统筹共享管理技术设计

针对航空摄影统筹共享管理缺乏有效技术手段的问题,在分析当前技术现状和技术要求的基础上,提出了建立综合管理系统开展航空摄影“事前-事中-事后”全流程在线统筹共享管理的技术思路,设计了航摄统筹共享管理的工作流程、系统组成、数据流与接口,详细阐述了“事前”开展航摄需求技术统筹;“事中”进行航摄信息可视化在线共享;“事后”实现航摄成果目录信息快速更新等技术内容。研究成果可为实现全国航空摄影“一获多用”与信息及时共享提供新的解决方案。

DIP付费对医疗服务项目诊疗强度的影响研究

目的:探索DIP付费对医疗服务项目诊疗强度的影响。方法:收集某市2020年1月至2022年12月330家医疗机构住院患者开展数字化摄影诊疗的数据,利用独立样本t检验比较诊疗强度的变化,采用间断时间序列分析考察DIP付费对诊疗强度的影响。结果:DIP付费后三级医院、一级医院数字化摄影诊疗强度分别增加了0.03和0.04,二级医院下降了0.06;三级医院诊疗强度在短期内增加(β_(2)=0.061),长期却呈现下降趋势(β_(3)=-0.007)。结论:DIP付费对医疗服务项目诊疗强度的政策效应可能与政策预期不一致,异质性和同质性并存;要强化DIP支付方式改革与医疗服务价格改革的协同路径。

面向水库型小流域的多源水利三维建模方法研究

针对大型河道三维可视化建模中采用激光雷达、声呐等传感器获取水下地形建模方式操作复杂且成本较高的缺陷,在利用基础测绘成果构建实景地形模型的基础上,采用多源三维模型融合方法对月塘水库小流域进行实景建模。首先采用倾斜摄影三维建模技术对重点水利工程进行实景建模,并利用专业建模软件对重要控制单元进行构件级建模;然后利用已有河道断面数据结合河流面状矢量数据,实现流域内胥浦河河道实景三维建模;最后通过人机交互方式对多源实景模型进行微调与相互印证,实现月塘水库小流域多源三维模型精准融合构建。研究成果可为河道型水库的预报调度与洪水风险分析提供可视化模型基础与决策支撑。

定位装置对头面部医学摄影一致性的影响

目的:探讨定位装置对整形美容外科头面部医学摄影一致性的影响。方法:于2021年3月,选取志愿者20例作为研究对象,采用自身对照研究法,每位志愿者先使用传统拍摄法拍摄三组照片,每组照片拍摄角度为左、右45°和左、右90°;30 min后,每位志愿者再使用定位装置拍摄法拍摄三组照片,记录每组拍摄时间。通过Image J软件对每组照片进行距离及角度的测量,对两种拍摄方法测量数据的差值及所用时间进行统计学分析。结果:定位装置拍摄法中左、右45°照的角度差,左、右90°照的角度差,以及左、右45°照的距离差均小于传统拍摄法,拍摄时间也明显短于传统拍摄法(P<0.05)。结论:摄影定位装置的运用,大大提升了整形美容外科头面部医学摄影的一致性,使影像资料更加标准、规范,还缩短了拍照时间,提升了工作效率,提高医、护、患三方满意度。

20°上臂倾斜位摄片法在肱骨外髁骨折中的应用价值

目的浅析20°上臂倾斜位摄片法在肱骨外髁骨折中的应用价值。方法选取2022年6月至2023年5月江西省儿童医院收治的50例肱骨外髁骨折患儿作为研究对象,所有患儿均采用传统正位与20°上臂倾斜位摄片法,手术结果作为金标准。并根据Jakob分类法进行二次评估,测量不同体位X线正位平片中骨块内侧与外侧的移位距离。结果50例肱骨外髁骨折患儿中,标准位摄片法的Jakob分型为:Ⅰ度29例、Ⅱ度17例、Ⅲ度4例;20°上臂倾斜位摄片法的Jakob分型为:Ⅰ度19例、Ⅱ度27例、Ⅲ度4例。经过测量可知,20°上臂倾斜位摄片法的异位距离大于标准位摄片法,差异有统计学意义(P<0.05)。20°上臂倾斜位摄片法准确率高于标准位摄片法,差异有统计学意义(P<0.05)。50例肱骨外髁骨折患儿中,手术结果为金标准,45例结果正确,5例结果错误;20°上臂倾斜位摄片法灵敏度为97.73%、特异度为66.67%、准确度为94.00%。结论在肱骨外髁骨折中采用20°上臂倾斜位摄片法,能够快速显示患儿肱骨外髁骨折的骨块移位程度,提高诊断的准确率,为临床治疗提供有利的参考依据。

柱形空泡动力学研究综述

空化现象是核工程领域的重要研究内容,而柱形空泡动力学是其前沿方向。本文从理论和实验两个方面对柱形空泡的研究进展进行了综述。首先,本文从流体力学基本方程出发,对三种典型形式的柱形空泡动力学方程进行了详细的推导和系统的整理。然后,对这三种柱形泡动力学方程进行了比较,指出了它们在方程的形式和适用范围上的差别。此外,文章分析了方程中一些重要参数对方程预测结果的影响。实验研究方面,本文展示了柱形空泡典型的高速摄影图片以及空泡半径随时间的变化。通过将实验结果与理论结果进行对比,发现二者具有良好的一致性。

基于倾斜三维模型与机载LiDAR点云的DEM快速更新

随着倾斜摄影技术的发展和实景三维建设的不断推进,地形级实景三维地理场景对DEM现势性要求日益提高。为了解决传统方法更新DEM成本高、工作量大的问题,本文提出了利用倾斜三维模型成果转换点云,融合两期DSM和DOM进行变化检测,替换已有机载LiDAR点云,快速更新DEM的方法,并通过项目验证了该方法的可行性,极大地节约了生产成本,为DEM更新提供了思路。

倾斜摄影三维模型中水域空洞的自动识别与修复

水域具有独特的光学反射特性,在倾斜摄影测量数据生成的模型中易导致空洞产生。为保证三维模型的完整性和显示效果,需要对水域空洞进行识别与修复。由于倾斜摄影模型内部存在大量独立的几何体,采用常规的三角网拓扑关系提取水域空洞轮廓线时会提取出所有几何体的边界轮廓线,导致无法准确识别水域空洞边界。针对此问题,本文提出了一种基于坐标转换和图像边缘检测获取水域空洞轮廓线的新方法:首先,根据OSGB模型中Z轴方向上是否存在三维点绘制水域空洞二值图,并对二值图像进行边缘检测获取水域空洞的二维轮廓线;再通过坐标转换获取水域空洞的三维轮廓线,对三维轮廓线进行约束构网、纹理映射,得到水域空洞的几何体并将绘制的水域几何体切割融合到不同的模型文件;最终实现水域空洞轮廓的自动识别以及水域空洞的自动修复。

3种典型炸药水中爆炸初始冲击波的近场参数测试

为了探索水中爆炸初始冲击波近场的传播特性和衰减差异,扩展水箱法测爆压试验的研究范围,采用高速摄影狭缝扫描测试方法,对3种典型的柱状炸药(TNT、JH-14、JO-9159)水中爆炸过程进行了测试研究,测量获得了一维正平面水中初始冲击波传播的时间历程演化曲线;由此拟合计算出柱状炸药的水中初始冲击波传播速度随时间的变化曲线,进一步由水的冲击雨果尼奥方程和冲击波动量方程、界面上压力和质点速度连续的条件及声学近似理论,计算得到了水中爆炸初始冲击波阵面压力的衰减曲线和炸药试样的爆压、多方指数等爆轰参数。结果表明,水中爆炸初始冲击波阵面压力近似以指数形式衰减,初始阶段衰减极快;当炸药爆压为20~37 GPa时,水中冲击波初始速度处于5.66~6.81 km/s的范围,波阵面初始压力与炸药的爆压近似成线性关系;虽然JO-9159的爆速和爆压均高于JH-14和TNT,但其水中初始冲击波传播仅在7μs后,其冲击波速度与波阵面压力下降至最低,说明以HMX为基的理想炸药水中爆炸初始冲击波速度和波阵面压力的衰减率要高于以RDX为基的理想炸药和TNT,高爆速、高爆压类理想炸药在水中初始冲击波性能方面优势不显著。