Vehicle Head and Tail Recognition Algorithm for Lightweight DCDSNet

In the model of the vehicle recognition algorithm implemented by the convolutional neural network,the model needs to compute and store a lot of parameters.Too many parameters occupy a lot of computational resources making it difficult to run on computers with poor performance.Therefore,obtaining more efficient feature information of target image or video with better accuracy on computers with limited arithmetic power becomes the main goal of this research.In this paper,a lightweight densely connected,and deeply separable convolutional network(DCDSNet)algorithmis proposed to achieve this goal.Visual Geometry Group(VGG)model is improved by utilizing the convolution instead of the fully connected module,the deeply separable convolution module,and the densely connected network module,with the first two modules reducing the parameters and the third module allowing the algorithm to have more features in a limited number of parameters.The algorithm achieves better results in the mine vehicle recognition dataset.Experiments show that the recognition accuracy is improved by 4.41% compared to VGG19 and the amount of parameters is reduced by 71% compared to VGG19.

Unmanned wave glider heading model identification and control by artificial fish swarm algorithm

We introduce the artificial fish swarm algorithm for heading motion model identification and control parameter optimization problems for the“Ocean Rambler”unmanned wave glider(UWG).First,under certain assumptions,the rigid-flexible multi-body system of the UWG was simplified as a rigid system composed of“thruster+float body”,based on which a planar motion model of the UWG was established.Second,we obtained the model parameters using an empirical method combined with parameter identification,which means that some parameters were estimated by the empirical method.In view of the specificity and importance of the heading control,heading model parameters were identified through the artificial fish swarm algorithm based on tank test data,so that we could take full advantage of the limited trial data to factually describe the dynamic characteristics of the system.Based on the established heading motion model,parameters of the heading S-surface controller were optimized using the artificial fish swarm algorithm.Heading motion comparison and maritime control experiments of the“Ocean Rambler”UWG were completed.Tank test results show high precision of heading motion prediction including heading angle and yawing angular velocity.The UWG shows good control performance in tank tests and sea trials.The efficiency of the proposed method is verified.

Dissimilar survival and clinicopathological characteristics of mucinous adenocarcinoma located in pancreatic head and body/tail

BACKGROUND Growing evidence shows that pancreatic tumors in different anatomical locations have different characteristics,which have a significant impact on prognosis.However,no study has reported the differences between pancreatic mucinous adenocarcinoma(PMAC)in the head vs the body/tail of the pancreas.AIM To investigate the differences in survival and clinicopathological characteristics between PMAC in the head and body/tail of pancreas.METHODS A total of 2058 PMAC patients from the Surveillance,Epidemiology,and End Results database diagnosed between 1992 and 2017 were retrospectively reviewed.We divided the patients who met the inclusion criteria into pancreatic head group(PHG)and pancreatic body/tail group(PBTG).The relationship between two groups and risk of invasive factors was identified using logistic regression analysis.Kaplan-Meier analysis and Cox regression analysis were conducted to compare the overall survival(OS)and cancer-specific survival(CSS)of two patient groups.RESULTS In total,271 PMAC patients were included in the study.The 1-year,3-year,and 5-year OS rates of these patients were 51.6%,23.5%,and 13.6%,respectively.The 1-year,3-year,and 5-year CSS rates were 53.2%,26.2%,and 17.4%,respectively.The median OS of PHG patients was longer than that of PBTG patients(18 vs 7.5 mo,P<0.001).Compared to PHG patients,PBTG patients had a greater risk of metastases[odds ratio(OR)=2.747,95%confidence interval(CI):1.628-4.636,P<0.001]and higher staging(OR=3.204,95%CI:1.895-5.415,P<0.001).Survival analysis revealed that age<65 years,male sex,low grade(G1-G2),low stage,systemic therapy,and PMAC located at the pancreatic head led to longer OS and CSS(all P<0.05).The location of PMAC was an independent prognostic factor for CSS[hazard ratio(HR)=0.7,95%CI:0.52-0.94,P=0.017].Further analysis demonstrated that OS and CSS of PHG were significantly better than PBTG in advanced stage(stage III-IV).CONCLUSION Compared to the pancreatic body/tail,PMAC located in the pancreatic head has better survival and favorable clinicopathological characteristics.

曲线多尺度表达的Head-Tail信息量分割法

本文提出一种基于Head-Tail信息量分割的地理要素多尺度表达模型。首先通过傅里叶变换将地理线要素转换为傅里叶描述子,并通过香农信息熵理论计算其频域信息量。然后,按Head-Tail数据分布模式确定地理要素信息量的分界点,并设计函数对各个分界点所对应的傅里叶截断频率进行估计。最后,参考传统方根模型,建立以地理要素频率信息量为基础的信息方根模型,计算与各个地图层次相对应的关键尺度,实现地理要素的层次化多尺度表达。采用等高线及海岸线的数据试验表明,本文所提出的模型能够有效根据设定的比例尺对地理要素进行化简,对不同目标比例尺的简化结果体现出了良好的区分度与层次性。同时,在保证化简结果与原地理要素面积重叠比一致的情况下,本文模型所得到的结果优于传统的简化算法。

Heuristic algorithms for scheduling on uniform parallel machines with heads and tails

This paper considers the uniform parallel machine scheduling problem with unequal release dates and delivery times to minimize the maximum completion time.For this NP-hard problem,the largest sum of release date,processing time and delivery time first rule is designed to determine a certain machine for each job,and the largest difference between delivery time and release date first rule is designed to sequence the jobs scheduled on the same machine,and then a novel algorithm for the scheduling problem is built.To evaluate the performance of the proposed algorithm,a lower bound for the problem is proposed.The accuracy of the proposed algorithm is tested based on the data with problem size varying from 200 jobs to 600 jobs.The computational results indicate that the average relative error between the proposed algorithm and the lower bound is only 0.667%,therefore the solutions obtained by the proposed algorithm are very accurate.

Comparative Analysis of Amino Acid Composition in the Head, Muscle and Tail of Fresh African Cat Fish (Clarias gariepinus)

Consumption of seafood has increased globally over the past 50 years, from an average of 9.9 kg per capital in the 1960s to over 20 kg in 2016. In Nigeria, African catfish (Clarias gariepinus) has gained more attention in terms of farming and consumption compared to other seafood. This project was carried out to determine the level of amino acid composition in the head, muscle and tail of catfish so as to enlighten the public about their choice on the part of catfish to be eaten. Eight different sets of fresh catfish with a mean ± standard deviation of 15.6 kg ± 1.89, which were collected from a local fish pound in Ado-Ekiti, Nigeria, were used for the analysis. Extraction and evaluation of the amino acid composition were carried out using standard analytical techniques. The results obtained showed that high values of amino acids were observed in the fresh catfish (g/100g protein): 13.27 - 13.87 (Glu), 8.76 - 10.39 (Asp), 7.20 - 9.15 (Leu), 7.00 - 8.26 (Lys) and a low value in (Cys), 1.27 - 1.38. The total amino acids had a range of 89.16 - 92.12 g/100g. The coefficient of variation percent (CV %) for all the parameters determined were all generally low with values of 0.74 (observed for Ser) and 39.41 (observed for Pro). Total essential amino acid with histidine values ranged from 38.64 - 42.91 g/100g (CV % = 5.66). It was also observed that the concentrations (g/100g) of amino acids in the muscle of fresh catfish (92.12) are higher than that of the tail (91.98) and that of the head (89.16). There is no appreciable variation in the amino composition of the head, muscle and tail of catfish as the values observed were found to be statistically insignificant (p > 0.05). Therefore, it could be posited that the amino acid in the muscle, tail and head of African catfish would therefore be enough to prevent malnutrition in children and in adults who feed solely on any part of this fish as a main source of protein.

英汉语中“head（首／头）、tail（尾）”隐喻研究

隐喻是一种利用已知事物来理解未知事物的认知手段，人类以自身结构去认识世界的人体隐喻也是隐喻中的一种常见方式，而人体中首、尾也是经常被用来指代时间、空间等概念。本文通过对比英汉两种语言中head（苔／头）和tail（尾）隐喻的不同，揭示了其跨语言、跨文化的共性及个性，在认知语言学和跨文化交际研究中都具有重要意义。

当归不同药用部位的化学成分及药理作用研究进展

当归作为“分根梢”理论的代表药材,其以不同药用部位分别入药起源于唐代,形成于金元,兴盛于明清;经历代医家临床实践总结,逐渐确立了归头止血、归身补血、归尾活血的作用功效。现代研究表明,当归中含有当归多糖、挥发油、有机酸、无机盐等药效部位及阿魏酸、绿原酸、欧前胡素、藁本内酯等药效成分;在当归头、当归身及当归尾等不同药用部位中,上述药效物质基础的种类分布均匀、区别不大,但含量及相关比例差异明显;多数学者认为,含量、比例差异可能是导致当归不同药用部位存在较大药理作用差异的主要原因。通过对相关文献的检索,梳理和总结近年来国内外对当归不同药用部位中药效物质基础分布情况及相关药理作用的研究,从当归不同药用部位中药效物质基础的种类分布和含量差异两个方面综合分析,并结合相关药效物质基础对应的药理作用研究,找寻当归不同药用部位间存在的“成分-效用”关系,以期为历代医家将当归按当归头、当归身、当归尾分别入药提供更强有力的佐证,同时为当归提出进一步研究的主要方向和思路。

不结球白菜抗根肿病种质资源的创制和抗性鉴定

以抗根肿病大白菜和不结球白菜为试材进行杂交,通过连续五代回交转育和自交分离纯化,创制了抗根肿病不结球白菜新种质2份,编号为BC20和Q20,并进行了比较试验。结果表明:创制的2份新种质综合农艺性状接近不结球白菜;属于中箕类型;生长势强;叶片绿色,宽椭圆;叶柄匙形,中长较窄,浅绿;BC20叶面较平整,Q20叶面略皱。该种质的创制为进一步培育抗根肿病不结球白菜新品种奠定了资源基础。

基于改进Yolov5s的增强现实文物识别方法

将增强现实技术应用于数字博物馆的文物展览,有助于拉近参观者与文物的距离,使展览更具趣味性。针对增强现实技术在文物展览应用场景中,增强现实设备采集的文物目标图像背景复杂、文物形状纹理丰富而导致的误检、识别准确率低的问题,提出一种基于改进的Yolov5s的文物识别方法。在Yolov5s网络结构中的骨干网络与颈部网络之间引入了CBAM注意力机制,并在骨干网络中的Bottleneck模块中,使用多头注意力机制替换普通卷积,有效捕获局部信息,降低了无用信息的干扰。为了提高识别网络对于目标文物的边界框定位精度,采用DIoU-NMS方法挑选最优的目标识别框作为最终的预测框。实验结果表明,该方法提高了文物的平均识别精度,比原模型更适用于AR应用文物的目标识别。

基于增强多头注意力机制的Optuna-BiGRU测井岩性识别

测井岩性识别是油气勘探开发中至关重要的内容。针对现有算法模型在处理测井曲线数据时,无法有效捕获曲线内部深层关联和深度方向关系、拟合能力较弱、难以准确提取关键特征、噪声干扰以及模型超参数调优过程复杂困难等问题,提出了一种通过Optuna超参数优化双向门循环单元(Optuna-BiGRU)结合增强多头注意力机制(EMHA)的测井岩性识别模型——Optuna-BiGRU-EMHA模型。该模型引入残差机制和层归一化以改进多头注意力机制模块,并结合双向门循环单元(BiGRU)解决了处理测井数据时的问题,同时使用Optuna超参数优化框架和小波包自适应阈值方法分别解决了超参数调优和噪声干扰问题。首先通过交会图分析和敏感性箱线图分析选取自然伽马、深感应电阻率、中子-密度孔隙度、平均中子-密度孔隙度和岩性密度5个特征参数的测井数据,通过小波包自适应阈值方法对数据进行去噪,并将测井数据分割成数据块,然后利用Optuna框架优化BiGRU-EMHA模型超参数,最后通过实验对比K-近邻算法(KNN)、随机森林(RF)、极端梯度提升算法(XGBoost)、长短期记忆(LSTM)神经网络、BiGRU、双向长短期记忆(BiLSTM)神经网络、BiGRU-MHA、Optuna-BiGRU-EMHA等8种模型在测井岩性识别中的精度。结果表明:Optuna-BiGRU-EMHA模型识别准确率达到80%,相对于传统机器学习模型和深度学习模型,综合岩性识别准确率分别提高15.94%~23.14%和3.93%~15.94%,该模型为常规测井岩性识别提供了坚实的理论支持。

城市群网络联系的幂律分布规律与层级结构特征

城市群的网络联系存在幂率分布规律,在这一规律的影响下,城市网络是“有等级倾向的网络”,其网络结构必然表现为明显的层级特征。首先,基于城市引力模型测度城市之间的相互联系,实证检验城市群城市间联系强度的概率分布规律,证明城市网络联系符合幂律分布的特征,且具有标度不变性。城市群的联系网络为服从幂率分布的无标度网络。然后,基于头尾断裂法对城市群城市网络的层级结构进行了划分和比较分析。尽管不同城市群城市网络的规模、强度和发育程度存在较大差异,强关联关系的空间形态也有明显的方向性不同,但其网络结构都具有明显的“核心—边缘”层级特征。

表面应变重构的视觉识别方法

针对全断面岩石隧道掘进机刀盘在恶劣工况下的应变监测,本文提出了一种表面应变重构的视觉识别方法。通过图像中圆形标距中心点之间像素点数量的变化,得到标距的实际变形,并在此基础上结合柯西-格林公式进行表面应力重构。本文以隧道掘进机刀盘刀座和筋板2个关键位置为研究对象,设计了4330V和Q345-D共2种材料不同尺寸标准样件,搭建了基于视觉识别和基于应变片的2套测量系统,进行了拉伸载荷作用下的应变对比及样件表面应力重构。研究表明:本文基于视觉识别方法测得的2种材料标准件变形的误差范围为1.28%~3.23%,在此基础上,进行的材料表面应力重构的误差范围为2.20%~4.76%。本文方法可以实现变形的准确测量及实现多方向的应变和变形重构,为实现复杂结构的应变和变形测量奠定了坚实基础。

融合GAT与头尾标签的多标签文本分类模型

现有基于神经网络的多标签文本分类主要存在两方面的不足,一是现有的数据中标签-文本频率存在长尾分布,二是很少从图结构中获取全局标签之间的关联性,针对以上问题,提出了一种融合GAT与头尾标签分类器的多标签文本分类模型(GATTN),该模型利用带有注意力机制的Bi-LSTM得到文本的特征向量表示;同时把不同标签之间的关联性转换为包含了全局信息的边加权图,利用多层的图注意力网络来学习头标签之间的关联性。然后将其与文本上下文语义信息进行点积交互,得到具有文本语义和头标签信息的特征表示提高模型的泛化能力。在AAPD、RCV1-V2和EUR-Lex三个公开英文数据集上的实验结果证明,该模型针对数据长尾分布的多标签文本分类效果优于其他基线模型。

融合外部知识图谱的多模态知识共创价值识别研究

[目的/意义]快速且精准地识别海量多模态数据中的价值性内容,对于促进知识传播、提升产出质量具有重要的意义。[方法/过程]基于可解释性视角聚焦知识共创中“用户+多模态知识”的双重推进机制。首先,依托BERT+BiLSTM与ResNet模型分别提炼文本与图片特征以获取多模态知识向量表示;其次,依据社会认知理论剖析用户行为,采用DeepFM捕捉交互特征间的关联生成用户向量表示;再次,借助K-BERT对文本数据嵌入知识图谱得到外部知识向量表示;最后,基于多头注意力机制融合各维度特征向量,通过动态调整权重完成价值内容的识别。[结果/结论]通过使用魅族Flyme社区数据进行实验,所构建的融合模型准确率达到88.31%,相较于其他基线模型与组合模型,评价指标均有不同程度的提升,证明嵌入外部知识并融合文本、图片与用户属性可以有效提升价值的识别效果。

顾及空间异质性的人口分布数据精细划分方法

针对公开人口分布数据难以精确刻画小区域内人口分布特征及其在同种地类不同区域存在人口差异性等问题,提出了一种顾及空间异质性的人口分布数据精细划分方法。该方法的核心思想是以格网为基础单元,分区建立人口与辅助数据间的关联,从而构建人口分布精细划分模型。采用首尾分割法实现分区策略,利用人口密度约束地类权重,在地类权重、地类人口密度权重与面积权重三重约束下对人口重新分配;在此基础上,利用Dasymetric法对优化的人口分布数据进行精细划分。结果表明,该方法的平均相对误差为5.71%,决定系数为0.98,不仅提高了人口分布数据的精度和分辨率,同时有效解决了当前公开人口分布数据存在的问题,打破了对人口统计数据的依赖,为人口精细化研究提供了新的思路。

基于Swin Transformer目标全景分割的三峡库首土质滑坡识别

【目的】滑坡识别是解决山区地质灾害隐患在哪里的关键。尤其人工智能是深度学习方法开始被广泛应用于目标识别领域,但对于多植被山区复杂环境下的滑坡隐患识别,存在着模型单一、精度较差等问题。【方法】故文章提出一种基于Swin Transformer(Shift Windows Transformer)作为骨干网络结合目标全景分割的智能识别方法,对三峡库首区域土质滑坡开展识别。将三峡库首的485处土质滑坡制作成样本集,并分为训练集和测试集。将训练集加载进Swin Transformer模型中进行训练,模型采用自注意力机制对训练集提取特征,构建特征图,测试集验证特征图的识别精度,保留识别精度最高的特征图。最终以此实现滑坡目标与背景区域的有效区分进而完成隐患识别,同时与DeepLab V3模型进行对比。【结果】结果显示:Swin Transformer模型在识别精度和识别速度上都要高于DeepLab V3模型,在三峡库首的试验中准确率可以达到83.55%,单张图片预测时间为0.18 s。【结论】结果表明:该方法能够在多植被山区复杂环境下快速识别土质滑坡,可为多植被山区的滑坡灾害调查提供参考。

基于首尾和全过程方法的中江县灌溉水有效利用系数测算分析

【目的】对比分析2种灌溉水有效利用系数(IWEUC)测算方法,研究灌区用水损失的主要环节。【方法】以四川省中江县为例,在全县范围内划定了4处分区(TW、TH、ON、FO分区),分别选取典型渠系和典型田块,利用动水测定法、经验公式法和田间灌溉试验进行IWEUC全过程测算分析;统计渠首引水量和其他补充水源供水量得到区域毛灌溉用水总量,并观测作物净灌溉用水量,进行首尾测算分析,与全过程测算结果进行对比。【结果】中江县TW分区渠系水利用系数最大为0.596,TH分区渠系水利用系数最小为0.564,加权平均值为0.580;北部的ON、TW分区由于工程配套良好,其田间水利用系数接近0.90,而TH、FO分区受地形等影响,田间水利用系数在0.85左右;在灌溉保证率75%条件下,全过程测算分析法得到IWEUC结果为0.503,首尾测算分析法得到结果为0.492,二者相对偏差为2.2%。【结论】中江县目前的灌溉水有效利用系数低于全国平均水平,还有较大的节水潜力。

八段锦摇头摆尾式联合PNF运动模式对脑卒中患者平衡功能的影响

目的:观察八段锦摇头摆尾式联合PNF颈部运动模式对脑卒中患者平衡功能的影响。方法:筛选60例符合纳入标准的脑卒中患者,采用随机数字表法分为对照组和观察组各30例。对照组患者接受常规康复治疗,观察组在对照组基础上联合运用八段锦摇头摆尾式及PNF颈部运动模式,两组患者均接受治疗3周。分别于治疗前后进行脑卒中患者姿势评定量表(PASS)评分、躯干损伤量表(TIS)评分、Berg平衡量表(BBS)评分评定。结果:治疗3周后,两组PASS评分、TIS评分、BBS评分均较治疗前明显提高(P<0.01),且观察组评分改善较对照组更明显(P<0.05)。TIS评分分项中静态坐位平衡及动态坐位平衡观察组与对照组改善差异有统计学意义(P<0.05),协调分项观察组与对照组改善差异无统计学意义(P>0.05)。结论:八段锦摇头摆尾式与PNF颈部运动模式相结合,可有效提升脑卒中患者姿势控制及平衡功能,较常规治疗展现出更佳的疗效。

基于首尾冗余的列控车载设备测速测距系统

测速测距系统主要任务是向列车控制系统提供实时的速度和位移信息,常见的测速测距系统主要采用多传感器(如速度传感器、雷达传感器、加速度计、卫星等)融合的方式。从系统的角度提出一种基于多种传感器融合的首尾冗余测速测距系统设计方案,既能有效地降低因单点故障导致的测速测距系统故障,又能避免因测速测距系统故障而导致的信号系统故障,从而实现了车头、车尾测速测距系统的安全冗余。