基于S-YOLO V5和Vision Transformer的视频内容描述算法

视频内容描述的自动生成是结合计算机视觉和自然语言处理等相关技术提出的一种新型交叉学习任务。针对当前视频内容生成描述模型可读性不佳的问题,本研究提出一种基于S-YOLO V5和Vison Transformer(ViT)的视频内容描述算法。首先,基于神经网络模型KATNA提取关键帧,以最少帧数进行模型训练;其次,利用S-YOLO V5模型提取视频帧中的语义信息,并结合预训练ResNet101模型和预训练C3D模型提取视频静态视觉特征和动态视觉特征,并对两种模态特征进行融合;然后,基于ViT结构的强大长距离编码能力,构建模型编码器对融合特征进行长距离依赖编码;最后,将编码器的输出作为LSTM解码器的输入,依次输出预测词,生成最终的自然语言描述。通过在MSR-VTT数据集上进行测试,本研究模型的BLEU-4、METEOR、ROUGEL和CIDEr分别为42.9、28.8、62.4和51.4;在MSVD数据集上进行测试,本研究模型的BLEU-4、METEOR、ROUGEL和CIDEr分别为56.8、37.6、74.5以及98.5。与当前主流模型相比,本研究模型在多项评价指标上表现优异。

Design of Online Vision Detection System for Stator Winding Coil

The quality of the stator winding coil directly affects the performance of the motor.A dual-camera online machine vision detection method to detect whether the coil leads and winding regions were qualified was designed.A vision detection platform was designed to capture individual winding images,and an image processing algorithm was used for image pre-processing,template matching and positioning of the coil lead area to set up a coordinate system.After eliminating image noise by Blob analysis,the improved Canny algorithm was used to detect the location of the coil lead paint stripped region,and the time was reduced by about half compared to the Canny algorithm.The coil winding region was trained with the ShuffleNet V2-YOLOv5s model for the dataset,and the detect file was converted to the Open Neural Network Exchange(ONNX)model for the detection of winding cross features with an average accuracy of 99.0%.The software interface of the detection system was designed to perform qualified discrimination tests on the workpieces,and the detection data were recorded and statistically analyzed.The results showed that the stator winding coil qualified discrimination accuracy reached 96.2%,and the average detection time of a single workpiece was about 300 ms,while YOLOv5s took less than 30 ms.

基于改进YOLOX-s算法的航天太阳电池缺陷检测

针对航天太阳电池表面缺陷检测问题,提出基于机器视觉与深度学习的缺陷检测方法。通过航天太阳电池缺陷检测系统获取图像,并依据企业电池片缺陷的分类标准构建航天太阳电池缺陷数据集。采用切片技术获取包含缺陷目标的子图像数据集,解决卷积和下采样操作信息丢失而导致召回率低的问题。针对不同缺陷采取适当的图像增强方式进行扩充数据集,以避免训练过程中因数据集不足导致的过拟合问题。采用深度可分离卷积、优化损失函数、双线性插值上采样及引入注意力机制等方法对YOLOX-s算法进行改进,以获得综合效果最佳的航天太阳电池缺陷检测模型。通过不同数据集训练及检测精度指标对比,以及消融实验验证改进模型的有效性。通过改进模型与同类主流模型对比实验,验证改进模型在航天太阳电池缺陷检测方面的优越性。

GAN-DIRNet:A Novel Deformable Image Registration Approach for Multimodal Histological Images

Multi-modal histological image registration tasks pose significant challenges due to tissue staining operations causing partial loss and folding of tissue.Convolutional neural network(CNN)and generative adversarial network(GAN)are pivotal inmedical image registration.However,existing methods often struggle with severe interference and deformation,as seen in histological images of conditions like Cushing’s disease.We argue that the failure of current approaches lies in underutilizing the feature extraction capability of the discriminator inGAN.In this study,we propose a novel multi-modal registration approach GAN-DIRNet based on GAN for deformable histological image registration.To begin with,the discriminators of two GANs are embedded as a new dual parallel feature extraction module into the unsupervised registration networks,characterized by implicitly extracting feature descriptors of specific modalities.Additionally,modal feature description layers and registration layers collaborate in unsupervised optimization,facilitating faster convergence and more precise results.Lastly,experiments and evaluations were conducted on the registration of the Mixed National Institute of Standards and Technology database(MNIST),eight publicly available datasets of histological sections and the Clustering-Registration-Classification-Segmentation(CRCS)dataset on the Cushing’s disease.Experimental results demonstrate that our proposed GAN-DIRNet method surpasses existing approaches like DIRNet in terms of both registration accuracy and time efficiency,while also exhibiting robustness across different image types.

Intersection capacity based on driver's visual characteristics

In order to reflect the influence of the drivers＇ characteristic differences on intersection capacity under a mixed traffic flow, a driver correction coefficient for the intersection capacity calculation according to the driver＇s visual characteristics is proposed. First, the parameters of the driver＇s visual characteristics at some real roads, including gaze fixation distribution, mean fixation duration, visual angle distribution and some other parameters at intersections, are collected. Then, the relationship between the traffic flow rate at intersections and the parameters of driver eye movements are established. The analytical results indicate that when the traffic flow is unsaturated, the parameters of driver eye movements change relatively little; however, when the traffic flow is saturated, the parameters of driver eye movements change drastically. Finally, the saturation-flow-rate model is modified according to the parameters of driver eye movements; thus, a capacity model of intersections considering the driver＇s visual characteristics is obtained.

Night Vision Characteristics and Application of MTV 2821CB CCD Camera

Aim To describe accumulating frames characteristics of CCD camera as a night vision detector. Methods Utilizing CCD external trigger, computer video capture card and image processing software, the image accumulation was made. Results The detection of the static object image whose illuminance on the CCD FPA(focal plane array) was less than 3 7×10 -5 lx was realized and the image's resolution of 300?TV lines was achieved. Conclusion This experimental system can provide a kind of night vision device capable of detecting the static object at low light level and with low cost compared to an image intensifier.

基于改进Faster RCNN的驾驶员手持通话检测

针对现有驾驶员通话行为识别误判率较高的问题,提出一种基于改进Faster RCNN的驾驶员行为检测方法,对驾驶员的违规手持通话进行检测.介绍了针对区域建议网络(RPN)及其损失函数的优化策略,并在原始Faster RCNN上运用多尺度训练、增加锚点数量以及引入残差扩张网络的方法增强网络检测不同尺寸目标的鲁棒性.基于车载平台上采集的驾驶员行为图像,对文中提出的方法进行仿真试验.结果表明:RPN和Faster RCNN通过交替优化共享特征提取网络部分,实现高效的目标检测,相较于原始Faster RCNN,检测精确度提高了3.8%,对环境的适应性更强.

Advances in Vision-Based Lane Detection：Algorithms,Integration,Assessment,and Perspectives on ACP-Based Parallel Vision

Lane detection is a fundamental aspect of most current advanced driver assistance systems(ADASs). A large number of existing results focus on the study of vision-based lane detection methods due to the extensive knowledge background and the low-cost of camera devices. In this paper, previous visionbased lane detection studies are reviewed in terms of three aspects, which are lane detection algorithms, integration, and evaluation methods. Next, considering the inevitable limitations that exist in the camera-based lane detection system, the system integration methodologies for constructing more robust detection systems are reviewed and analyzed. The integration methods are further divided into three levels, namely, algorithm, system,and sensor. Algorithm level combines different lane detection algorithms while system level integrates other object detection systems to comprehensively detect lane positions. Sensor level uses multi-modal sensors to build a robust lane recognition system. In view of the complexity of evaluating the detection system, and the lack of common evaluation procedure and uniform metrics in past studies, the existing evaluation methods and metrics are analyzed and classified to propose a better evaluation of the lane detection system. Next, a comparison of representative studies is performed. Finally, a discussion on the limitations of current lane detection systems and the future developing trends toward an Artificial Society, Computational experiment-based parallel lane detection framework is proposed.

Vision Screening in the Pediatrician’s Office

Objective: To assess current practices, attitudes, and perceived barriers toward pediatric vision screening. Patients and Methods: A link to a 9-question survey was electronically distributed to a national sample of 6000 pediatricians through Medical Marketing Services Inc. Data were collected using Survey Monkey. Results: Email open rate was 11%;37% of those who opened the email responded (225 respondents). Over ninety percent of respondents perform some type of vision screening at least yearly, although age at which screening began varied, with two thirds of respondents instituting formal vision screening after three years. Fifty eight percent of respondents were either extremely unsatisfied, unsatisfied or only somewhat satisfied with their current screening method. Preferred methods of screening and confidence of pediatricians in their ability to detect pathology varied for children under versus over age three. The least frequently used methods for all age groups were autorefraction and photoscreening. The most commonly reported barriers to screening were inadequate training (48%), time required for exam (42%), and inadequate reimbursement (32%). Conclusions: Perceived barriers to vision screening in the pediatrician office have been previously identified, and photoscreening and autorefraction have been identified as a possible means to circumvent them. In spite of the addition of new procedural codes, pediatricians continue to report similar barriers to screening.

Exploring Challenges in Developing a Smart and Effective Assistive System for Improving the Experience of the Elderly Drivers

As the overall population ages, driving-related accidents and injuries, associated with elderly drivers, have risen. Existing research about elderly drivers mainly focuses on factual data collection and analysis, indicating the elderly＇s growing fatal accident rates and their different behaviours compared to younger drivers. However, few research has focused on design-led practical solutions to mitigate the elderly＇s growing fatal accidents, by consid- ering their usability and body conditions, afflicting the elderly, such as decreased vision, hearing, and reaction times. In this paper, first, current worldwide situations on growing fatal accident rates for elderly drivers is reviewed and the key impact factors are identified and discussed with regarding to usability and design trend in the automotive technology for elderly. Second, existing smart vehicle technology-based solutions to promote safe driving are explored and their pros and cons are discussed and anal- ysed. Most of solutions are not created by people with driving difficulties, which are caused by health problems most commonly afflicting the elderly. Thirdly, diverse design-led research activities are taken, such as a survey, observation, and interviews to gain new understanding of what kinds of driving problems elderly drivers have and demonstrate how new system concepts could be developed for the elderly＇s benefits. Finally, it is found that the elderly＇s low vision and late reaction are main factors causing their driving difficulties. Based on this finding, usable vehicle system design ideas have been proposed, by utilising facial expression sensing technology as a solution. The proposed solutions would ensure reducing both the elderly＇s driving problems and high fatal accident rates and provide a more enjoyable driving environment for the elderly population.

Gaze Behavior and Landscape Experience Characteristics of Expressway Drivers

Landscape is the visual impression formed in the mind of the aesthetic subject from what he sees at a certain viewpoint.A good viewpoint field is a place without visual barrier in the visual direction.Under the premise of analyzing aesthetic basis of expressway,starting from behavioral basis of dynamic vision,visual behavior characteristics of drivers at high speed are studied.On this basis,landscape experience tempo and rhythm determined by gaze locus under different linear characteristics are explored,and the impact of spatial alignment on gaze behavior and landscape experience is further analyzed.

基于改进YOLOv5s的火灾烟雾检测算法研究

为了解决火灾烟雾检测算法中存在的错检、漏检以及实时性差等问题,提出了一种基于YOLOv5s的火灾烟雾检测模型。首先,使用Ghost Convolution模块代替原YOLOv5s网络结构中的常规卷积模块,在保持相同性能的基础上,降低检测模型的计算成本、减少模型参数;其次,在原YOLOv5s模型骨干网络中加入Vision Transformer结构,减少对卷积神经网络的依赖性,同时提高获取全局和局部特征的能力;最后,引入Coordinate Attention注意力机制,有效地提取特征信息,进一步提高检测的准确率。实验结果表明,所提出的火灾烟雾检测模型参数减少17%,准确率提高0.73%,检测速度提升22.5%,可以满足实际场景下的火灾烟雾检测。

Children's Vision

Ｃｈｉｌｄｒｅｎ＇ｓＶｉｓｉｏｎＣｈｉｌｄｒｅｎ＇ｓｌｉｖｅｓａｒｅｍｕｃｈｍｏｒｅｃｏｌｏｒｆｕｌｔｈａｎａｄｕｌｔｓｕｓｕａｌｌｙｔｈｉｎｋ．ａｎｄｃｈｉｌｄｒｅｎ＇ｓｗａｙｓｏｆｔｈｉｎｋｉｎｇａｒｅｓｏｍｅｔｉｍｅｓｑｕｉｔｅｄｉｆｆｅｒｅｎ...

Modification of the Genetic Polymorphism of Glutathione-S-Transferase (GSTM1 and GSTT1) in Motorcycle Drivers Exposed to BTEX in Cotonou

The glutathione-S-transferase genes mainly the GSTM1 and GSTT1 alleles are responsible for the synthesis of detoxication enzymes that can remove toxic substances. The objective of this study is to seek changes in the genetic polymorphism of glutathione-S-transferase GSTM1 and GSTT1 in motorcycle drivers exposed to BTEX. Our study group consists of 60 motorcycle drivers including 30 professional and 30 non-professional. Blood samples were preleveled from the study population in the EDTA tubes and DNA was extracted by the phenol/chloroform method. The PCR technique was used to determine the presence or absence of genes. Our results showed that the percentage of GSTM1 null genotype has a statistically significant difference (P = 0.02), while the percentage of GSTT1 null genotype was non-significant (P = 0.76) between the two groups. The percentage of deletion of both genes is higher in professional than non-professional motorcycle drivers. Air pollution in Cotonou by BTEX seems to influence the deletion of the GSTM1 and GSTT1 genes at a higher percentage among professional than non-professional motorcycle drivers.

Spatio-temporal characteristics of regional sustainable economic growth drivers of China

Intellectual capital(IC)has become a universal performance indicator for the socioeconomic development of countries and regions.Based on a review ofnational intellectualcapital(NIC)and regional intellectualcapital(RIC)evaluation literature,we used the regional intellectualcapitalindicator(RICI)as a model for China’s RIC evaluation to indirectly understand China’s potential economic growth drivers.Specifically,we collected statisticaldata of31 provinces(including municipalities and autonomous regions)in China from 2004 to 2016to measure RICI and analyze its dynamic characteristics from temporal and spatial perspectives.In this paper,Delphi analysis was used to construct RICI model,and cluster analysisand exploratory spatial data analysis were used to analyze the temporal and spatial characteristics of RICI in China.The results showedthat RICI,which represents China’s overall economic growth drivers,increases annuallyand is consistent with economic developmentlevelduring the study period.Regarding the geo-spatial space,RICI follows the trend of“high in the east and low in the west”,gradually decreasing from eastern to western China.ForRIC structure,the shape of the radar chart of IC structure located in the eastern coastal areas is usually biased towards strong external relational capital,while that in western China is generally biased towards structural capital.For spatial correlation,China’s RICI has dependence on geographical adjacent space and economic space.Our research can provide policy suggestions for the sustainable development of regional economy from an IC perspective.

Quantification of Ride Comfort Using Musculoskeletal Mathematical Model Considering Vehicle Behavior

This research aims to quantify driver ride comfort due to changes in damper characteristics between comfort mode and sport mode,considering the vehicle’s inertial behavior.The comfort of riding in an automobile has been evaluated in recent years on the basis of a subjective sensory evaluation given by the driver.However,reflecting driving sensations in design work to improve ride comfort is abstract in nature and difficult to express theoretically.Therefore,we evaluated the human body’s effects while driving scientifically by quantifying the driver’s behavior while operating the steering wheel and the behavior of the automobile while in motion using physical quantities.To this end,we collected driver and vehicle data using amotion capture system and vehicle CAN and IMU sensors.We also constructed a three-dimensional musculoskeletal mathematical model to simulate driver movements and calculate the power and amount of energy per unit of time used for driving the joints and muscles of the human body.Here,we used comfort mode and sport mode to compare damper characteristics in terms of hardness.In comfort mode,damper characteristics are soft and steering stability is mild,but vibration from the road is not easily transmitted to the driver making for a lighter load on the driver.In sport mode,on the other hand,damper characteristics are hard and steering stability is comparatively better.Still,vibration from the road is easily transmitted to the driver,whichmakes it easy for a load to be placed on the driver.As a result of this comparison,it was found that a load was most likely to be applied to the driver’s neck.This result in relation to the neck joint can therefore be treated as an objective measure for quantifying ride comfort.

Visual resolution under photopic and mesopic conditions in patients with Sjogren's syndrome

AIM:To focus on different visual resolution tasks under photopic and mesopic conditions in Sjogren’s syndrome patients compared to age-matched healthy controls.METHODS:The visual resolution measurements included high and low visual acuities and contrast sensitivity functions.These tests were conducted under photopic and then mesopic conditions.Twenty-one Sjögren’s syndrome patients and 21 aged-matched healthy volunteers completed all the measurements in this study.RESULTS:Sjogren’s syndrome patients have greater impairment in contrast sensitivity than standardized visual acuity.This reduction was significant under the mesopic condition.Also,Sjogren’s syndrome patients treated with pilocarpine suffer more than patients without pilocarpine treatment under low light conditions.CONCLUSION:Sjogren’s syndrome patients shows greater impairment in different visual resolution tasks due to dry eye symptoms.

On Hawthorne＇ Black Vision： There Is Evil in Human Heart -- A Struc- turalist＇ s view of ＂Dr. Rappaccini＇ s Daughter＂

Nathaniel Hawthorne is regarded as the pioneer of short story in America and he is renowned for moral fables purifying the darkness of human beings."Dr. Rappaccini's Daughter"is one of the representative one. It tells the story of Dr. Rappaccini, with drive of morbid scientific zeal, raises his daughter as the bodily poison. The paper is intended to explore how Hawthorne delivers his black vision on human nature in light of the structuralism.

扫描仪 HP ScanJet 4S和Visioneer PaperPort Vx:性格迥异的双胞胎

Danny DeVito和Arnold Sch-warzenegger在电影Twins中扮演了一对性格各异的双胞胎。