A CNN-Based Single-Stage Occlusion Real-Time Target Detection Method

Aiming at the problem of low accuracy of traditional target detection methods for target detection in endoscopes in substation environments, a CNN-based real-time detection method for masked targets is proposed. The method adopts the overall design of backbone network, detection network and algorithmic parameter optimisation method, completes the model training on the self-constructed occlusion target dataset, and adopts the multi-scale perception method for target detection. The HNM algorithm is used to screen positive and negative samples during the training process, and the NMS algorithm is used to post-process the prediction results during the detection process to improve the detection efficiency. After experimental validation, the obtained model has the multi-class average predicted value (mAP) of the dataset. It has general advantages over traditional target detection methods. The detection time of a single target on FDDB dataset is 39 ms, which can meet the need of real-time target detection. In addition, the project team has successfully deployed the method into substations and put it into use in many places in Beijing, which is important for achieving the anomaly of occlusion target detection.

Highly Sensitive Ammonia Gas Sensors at Room Temperature Based on the Catalytic Mechanism of N,C Coordinated Ni Single-Atom Active Center

Significant challenges are posed by the limitations of gas sensing mechanisms for trace-level detection of ammonia(NH3).In this study,we propose to exploit single-atom catalytic activation and targeted adsorption properties to achieve highly sensitive and selective NH3 gas detection.Specifically,Ni singleatom active sites based on N,C coordination(Ni-N-C)were interfacially confined on the surface of two-dimensional(2D)MXene nanosheets(Ni-N-C/Ti_(3)C_(2)Tx),and a fully flexible gas sensor(MNPE-Ni-N-C/Ti_(3)C_(2)Tx)was integrated.The sensor demonstrates a remarkable response value to 5 ppm NH3(27.3%),excellent selectivity for NH3,and a low theoretical detection limit of 12.1 ppb.Simulation analysis by density functional calculation reveals that the Ni single-atom center with N,C coordination exhibits specific targeted adsorption properties for NH3.Additionally,its catalytic activation effect effectively reduces the Gibbs free energy of the sensing elemental reaction,while its electronic structure promotes the spill-over effect of reactive oxygen species at the gas-solid interface.The sensor has a dual-channel sensing mechanism of both chemical and electronic sensitization,which facilitates efficient electron transfer to the 2D MXene conductive network,resulting in the formation of the NH3 gas molecule sensing signal.Furthermore,the passivation of MXene edge defects by a conjugated hydrogen bond network enhances the long-term stability of MXene-based electrodes under high humidity conditions.This work achieves highly sensitive room-temperature NH3 gas detection based on the catalytic mechanism of Ni single-atom active center with N,C coordination,which provides a novel gas sensing mechanism for room-temperature trace gas detection research.

Target reliability index for serviceability limit state of single piles

The objective is to develop an approach for the determination of the target reliability index for serviceability limit state(SLS) of single piles. This contributes to conducting the SLS reliability-based design(RBD) of piles. Based on a two-parameter,hyperbolic curve-fitting equation describing the load-settlement relation of piles, the SLS model factor is defined. Then, taking into account the uncertainties of load-settlement model, load and bearing capacity of piles, the formula for computing the SLS reliability index(βsls) is obtained using the mean value first order second moment(MVFOSM) method. Meanwhile, the limit state function for conducting the SLS reliability analysis by the Monte Carlo simulation(MCS) method is established. These two methods are finally applied to determine the SLS target reliability index. Herein, the limiting tolerable settlement(slt) is treated as a random variable. For illustration, four load test databases from South Africa are compiled again to conduct reliability analysis and present the recommended target reliability indices. The results indicate that the MVFOSM method overestimates βsls compared to that computed by the MCS method. Besides, both factor of safety(FS) and slt are key factors influencing βsls, so the combination of FS and βsls is welcome to be used for the SLS reliability analysis of piles when slt is determined. For smaller slt, pile types and soils conditions have significant influence on the SLS target reliability indices; for larger slt, slt is the major factor having influence on the SLS target reliability indices. This proves that slt is the most key parameter for the determination of the SLS target reliability index.

Elaborate calibration procedure for cell irradiation at the CAS-LIBB single-particle microbeam

Single-particle microbeam is uniquely capable of precisely delivering a preset number of charged particles to individual cells or sub-cellular targets to be determined in vitro, It is crucial to find a reference point that relates the microbeam＇s location to the microscope＇s plane, and align individual targets at this reference point for cell irradiation. To choose an appropriate reference point, an approach based on analysing the intensity distribution of fluorescence in a thin scintillator excited by traversing particles is newly developed using the CAS-LIBB single-particle microbeam, which features decisive physical signification and sufficient resolution. As its bonus, this on-line analysis provides precise and fast response to the determination of beam profile and potentially optimizes the microbeam quality by further adjusting hardware setup.

Construction of SNP genetic maps based on targeted next-generation sequencing and QTL mapping of vital agronomic traits in faba bean(Vicia faba L.)

Owing to the limitation of a large genome size(~13 Gb),the genetic and gene mapping studies on faba bean(Vicia faba L.)are lagging far behind those for other legumes.In this study,we selected three purified faba bean lines(Yundou 8137,H0003712,and H000572)as parents and constructed two F2 populations.These two F2 populations,namely 167 F2 plants in Pop1(Yundou 8137×H0003712)and 204 F2 plants in Pop2(H000572×Yundou 8137),were genotyped using a targeted next-generation sequencing(TNGS)genotyping platform,and two high-density single nucleotide polymorphisms(SNP)genetic linkage maps of faba bean were constructed.The map constructed from Pop1 contained 5103 SNPs with a length of 1333.31 cM and an average marker density of 0.26 cM.The map constructed from Pop2 contained 1904 SNPs with a greater length of 1610.61 cM.In these two F2 populations,QTL mapping identified 98 QTLs for 14 agronomic traits related to the flowers,pods,plant types and grains.The two maps were then merged into an integrated genetic linkage map containing 6895 SNPs,with a length of 3324.48 cM.These results not only lay the foundation for fine mapping and map-based cloning of related genes,but can also accelerate the molecular marker-assisted breeding of faba bean.

Distance to Isocenter Directly Affects Margin and Inappropriate Margin Increases the Risk of Local Control Failure in LINAC-Based Single-Isocenter SRS or SRT for Multiple Brain Metastases

=In the advancement of single-isocenter multiple target treatment in the LINAC-based SRS or SRT, the target distance to the isocenter and grouping of multiple targets are the highly concerned and debatable topics in the SRS/SRT field at present. Three failure and success cases of local control in our early practices are presented in this study and it indicated that the target distance to the isocenter directly affects the margin and an inappropriate margin increase the risk of local control failure. The GTV expansion margin should be LINAC-specific and institute-specific. Within the physics and dosimetry scope, the AHARA (as high as reasonably achievable) principle is the first time proposed to the radiation oncology field. Radiobiology and tumor response complexity is beyond this study.

两种钛合金半穿甲战斗部侵彻单层钢靶性能试验研究

选取亚稳态β型TB17和高损伤容限α+β型TC4-DT两种高韧钛合金材料,对2种钛合金材料在高应变率条件下动态力学响应特性进行测试分析,并应用125 mm火炮发射装置开展2种钛合金半穿甲战斗部侵彻单层钢靶性能试验研究。结果表明,作为半穿甲战斗部的壳体材料,应选择屈服强度与冲击韧性匹配性好的钛合金,其中更应关注钛合金材料的塑性变形能力和抗冲击韧性;随着侵彻着角的增加,侵彻单层钢靶过程弹体头部侵蚀量及径向墩粗程度更严重,不利于弹体保持侵彻强度及装药安定性。

旋转物体的单光子三维重建技术

单光子探测技术具有高灵敏度和高时间分辨率,已成为三维成像领域的重要发展方向。APD单光子探测数据为表征光子脉冲事件的0-1数据,通常需要多次积累,可实现目标三维成像,但旋转物体的运动导致无法直接积累。针对旋转物体三维重建问题,提出了一种基于最大强度图清晰度的旋转物体单光子三维重建方法。首先建立旋转物体的三维重建模型,然后建立关于旋转速度的似然估计模型,最后以强度图的最大清晰度为代价函数求解旋转速度并进行重建。计算机仿真模拟了旋转十字架物体的单光子三维成像探测数据,通过三维重建算法对旋转物体实现了高精度三维重建,并在不同旋转速度下进行仿真验证,实现了不同速度下转速估计误差优于5%,深度重建误差优于0.1 m,验证了重建算法的正确性。

基于YOLOv4的行人检测算法

针对实际场景下YOLOv4模型难以处理遮挡行人的问题,在保证YOLOv4模型实时性的情况下做出了改进,将YOLOv4模型应用于行人检测。为了提高模型检测遮挡行人的能力,模型采用K-means++聚类算法重新设计适用于行人目标尺寸的先验框,引入排斥损失函数项,使候选框与临近的非匹配目标真实框距离最大化,候选框和其他目标真实框的重叠比例最小化。改进后模型在具有挑战性的数据集CrowdHuman和Caltech上进行实验,实验结果均验证了文中方法的有效性,最后将模型应用于实际场景下的视频行人检测,同样验证了本文改进的有效性。

基于繁殖策略的求解昂贵约束单目标进化算法

实际工程优化中存在大量约束优化问题,且有一些优化问题目标函数和约束函数的评价非常耗时,导致该类问题无法直接使用传统优化算法求解。为此,为了在评价次数有限的情况下获得较好的可行解,针对昂贵单目标约束优化问题,为评价费时的目标函数和约束函数建立径向基函数(Radial Basis Function,RBF)预测模型,以及根据估值自适应选择个体的繁殖策略,以期能产生较好的可行解。在7个标准测试函数及3个工业测试函数上的测试结果表明,相比于其它现有针对昂贵约束问题的优化方法,本方法无需确保初始种群中必须有可行解,且能在优化目标和约束函数评价次数有限的情况下找到更好的解。

基于深度学习的木材缺陷智能检测的研究进展与展望

木材作为天然生物材料很容易受到内外界影响从而产生不符合人们生产需求的缺陷,人们为了准确高效的识别木材缺陷进行了大量的研究。本文对近年来基于深度学习的木材缺陷检测技术进行梳理,根据使用方法的侧重点不同将其分类,并针对典型方法加以细分归类和对比分析,总结了每种方法的优缺点及其应用面。此外,提出了基于深度学习的木材缺陷检测技术目前所存在的难点与所陷困境。

Modeling of the Multi-Target Locating and Tracking in the Field Artillery System

A method for the multi target locating and tracking with the multi sensor in a field artillery system is studied. A general modeling structure of the system is established. Based on concepts of cluster and closed ball, an algorithm is put forward for multi sensor multi target data fusion and an optimal solution for state estimation is presented. The simulation results prove the algorithm works well for the multi stationary target locating and the multi moving target tracking under the condition of the sparse target environment. Therefore, this method can be directly applied to the field artillery C 3I system.

冬虫夏草及其近缘品中铅、镉、砷污染评价及人体健康累积风险评估方法探索

目的:对冬虫夏草及其近缘品中铅(Pb)、镉(Cd)、砷(As)的残留量进行测定,探索符合冬虫夏草及相关产品使用特点的污染评价及人体健康风险评估方法。方法:基于冬虫夏草及其近缘品中Pb、Cd、As的残留量监测数据,综合运用单因子污染指数法、尼梅罗综合指数法、金属污染指数法对冬虫夏草及其近缘品进行重金属污染评价,计算重金属日暴露量,分别采用危害指数法和更加精确的靶器官毒性剂量法对Pb、Cd、As联合暴露产生的健康风险进行累积风险评估。结果:污染评价结果说明,冬虫夏草及其近缘品中As的污染应引起关注,不同品种污染程度为冬虫夏草(繁育品)=蛹虫草<冬虫夏草(野生品)<香棒虫草<亚香棒虫草;人体健康风险评估结果表明,对于心血管和神经系统,1批冬虫夏草(野生品)全草中Pb、Cd、As联合暴露产生的累积健康风险需被进一步关注。结论:冬虫夏草及其近缘品重金属污染评价,以及人体健康风险评估方法,可为中药安全性评价及相关限量标准的制修订提供参考。

一种余光目标深度及高度的单帧图像测量

在自动避障,目标监控,自动驾驶等领域,基于单目视觉的目标测量总是关键技术之一.但是,已知的单目视觉测量方法只能得到目标的距离和最低点位置坐标,目标高度需要利用其他传感器获取.本文从目标的线特征出发,创造性的提出了关于余光目标像与像平面竖直中线的夹角、余光目标偏离单目相机装置的角度以及单目镜头的俯仰角三个变量之间的关系模型.最后,在单帧目标图像和单目相机的高度已知的情况下,利用该模型实现余光目标深度和高度的测量.实验结果表明,本文提出的测量方法能有效,快速的对目标的高度和距离进行测量.

海面目标光电单站无源定位与误差分析

为实现光电侦察设备对海面舰船目标单站无源定位,提出了对海面目标的高精度近实时光电单站无源定位方法。建立了单站无源定位方程,分析了目标定位误差,探讨了影响定位误差因素,给出了单站无源定位的应用建议。仿真结果表明:在17.899 km距离处单站无源定位的定位误差约为1.672%R,单次定位时间约20 ms;位置、纬度、经度和高度误差分别服从N(299.31 m,197.39 m)、N(4.944×10^(-5°),2.665×10^(-3°))、N(4.038×10^(-5°),1.899×10^(-3°))、N(-0.485 m,14.165 m)的正态分布;俯仰角、方位角和目标高度的误差对定位精度影响较大,俯仰角在[-120°,-15°]范围内定位结果可靠,目标高度误差在[-200 m,200 m]范围内,定位误差小于800 m,实现了对海面目标的高精度近实时单站无源定位。提出的方法适用于具备高度参考面估计的海面、地面和空中目标的单站无源定位。

寒喘祖帕单煎配方与配伍合煎对16HBE细胞抗炎活性及非靶向代谢组学差异的研究

目的考察寒喘祖帕按照传统配伍合煎及单煎配方的抗炎活性,并采用UPLC-Q-TOF/MS研究非靶向代谢组学差异。方法通过体外脂多糖诱导16HBE炎症细胞,使用CCK-8法筛选出单煎配方组和配伍合煎组的剂量;采用逆转录酶聚合酶链反应(Reverse Transcription-Polymerase Chain Reaction,RT-PCR)技术测定细胞中炎症因子IL-6、IL-8和TNF-α的mRNA表达水平,并采用UPLC-TOF/MS技术对细胞进行非靶向代谢组学分析:分为正常组(16HBE细胞+培养基)、模型组(16HBE细胞+培养基+0.1μg/mL LPS)、寒喘祖帕单煎配方组(16HBE细胞+培养基+0.1μg/mL LPS+50μg/mL单煎配方)、寒喘祖帕配伍合煎组(16HBE细胞+培养基+0.1μg/mL LPS+50μg/mL配伍合煎)。以此探究寒喘祖帕颗粒发挥抗炎活性的相关代谢通路,寻找寒喘祖帕颗粒抗炎活性相关的差异代谢物。结果与正常组比较,模型组在给药浓度为0.1μg/mL时细胞生长存活率未受到限制,差异无统计学意义(P>0.05);配伍合煎组在给药浓度为0~400μg/mL范围内时,细胞存活率未受影响,浓度>800μg/mL时,细胞生长存活率下降,差异具有统计学意义(P<0.01);而单煎配方组在浓度<50μg/mL时,细胞生长存活率未受到限制,浓度≥100μg/mL时,细胞生长存活率下降,差异具有统计学意义(P<0.01);RT-PCR实验结果显示:与正常组比较,模型组的IL-6、IL-8和TNF-αmRNA的表达上升,差异具有统计学意义(P<0.001),50μg/mL单煎配方较100μg/mL,IL-6、IL-8和TNF-αmRNA的表达下降,差异具有统计学意义(P<0.001);在相同药物浓度干预下,与正常组比较,模型组的IL-6、IL-8和TNF-αmRNA的表达上升,差异具有统计学意义(P<0.001),与模型组比较,单煎配方组和配伍合煎组可同时降低IL-6、IL-8和TNF-αmRNA的表达,差异具有统计学意义(P<0.001);配伍合煎组与单煎配方组比较,配伍合煎组IL-6、IL-8 mRNA的表达较单煎配方低,差异具有统计学意义(P<0.05)。代谢组学分析结果筛选出模型组与配伍合煎组间的内源性差异代谢物68个,主要与色氨酸代谢、甘油磷脂代谢和嘧啶代谢等有关;模型组与单煎配方组间识别出内源性差异代谢物62个,主要与甘油磷脂代谢、精氨酸和脯氨酸代谢等有关。结论寒喘祖帕单煎配方与配伍合煎对LPS诱导的炎症细胞模型抗炎效果存在差异,非靶向代谢组学分析结果表明两种方式通过不同的代谢通路发挥抗炎活性。

一种对高速运动目标的单光子测距方法

作为一种前沿的激光探测技术,单光子激光测距技术已成功应用于月球测距、卫星测距和地面测高等领域。然而,单光子测距在机载空对空、地对空平台上对高速运动目标进行跟踪测距时,回波光子落在不同的时间窗,导致直接计数无法有效提取信号的问题仍需解决。针对空对空条件下单光子激光测距的应用需求,基于时间相关光子计数技术设计一种适用于全天时、宽时域、多噪声条件下对高速运动目标的单光子测距方法。该方法采用阵列单光子探测器和相邻时间窗相关统计多帧处理算法提取激光回波光子信号,并在Matlab平台上对算法进行仿真实验,使用多元阵列单光子探测器实现最大测程百公里以上、背景噪声计数率约为5 MHz、单脉冲回波光子计数平均值为1条件下的回波光子信号提取。该方法能够克服传统单光子探测只能对准静态目标测距,只能在小接收视场和小波门范围等弱背景噪声及目标轨迹可预测条件下应用的限制,将单光子探测由只能固定平台夜晚对准静态目标测距推广至通用平台全天时对高速运动目标测距。

融合时空信息的Transformer单目标跟踪算法

目前,主流的基于孪生网络的单目标跟踪方法,通过计算模板与搜索区域之间的相似度来匹配目标,缺乏对目标时空状态信息的利用。特别是当场景中存在多个相似目标时,孪生网络跟踪器往往无法精确区分目标,从而导致跟踪错误。针对上述问题,提出一种融合时空信息的Transformer单目标跟踪算法(SIFTransT)。该算法通过MixFormer(end-to-end tracking with iterative mixed attention)跟踪器获取初步的跟踪结果,设计了一个目标状态计算模块,用于计算并存储目标的状态信息,包括目标位置、边界框、速度、加速度、运动方向等,以此深入挖掘目标状态信息。构建了一个基于Transformer的时空信息融合模块,利用编码器的自注意力和解码器的交叉注意力,深入融合目标最近一段时间的状态信息,从而更加准确地对目标状态进行建模,提高目标跟踪的准确性。在LaSOT数据集上的实验结果表明,相比基准算法MixFormer,SIFTransT算法在AUC指标提高了2.8个百分点,PNorm指标提升了2.6个百分点,P指标提升了2.1个百分点,在搭载RTX8000显卡的服务器上平均每秒处理帧数达28帧。

单细胞测序在口腔鳞状细胞癌研究中的应用

口腔鳞状细胞癌是一种常见口腔恶性肿瘤,易转移、复发和产生耐药性。单细胞测序及在此基础上发展而来的空间转录组技术可以获取单个细胞水平的转录组信息,能更加深入地研究细胞种群、功能、变化和通讯。本文从细胞分型、肿瘤微环境、肿瘤侵袭和转移、治疗靶点与预后标志、免疫治疗、肿瘤内微生物等方面对单细胞测序技术在口腔鳞状细胞癌研究中的应用进行归纳总结,并探讨当前研究的不足与未来改进方向。

基于体表标志定位急诊单引流管双靶点治疗丘脑出血破入脑室伴脑积水

目的 探讨颞部体表定位急诊单引流管双靶点丘脑血肿脑室引流术在丘脑出血破入脑室伴脑积水治疗中的应用价值。方法 回顾性选择丘脑出血破入脑室伴脑积水患者223例,其中包括采取颞部体表定位急诊单引流管双靶点丘脑血肿脑室引流术的研究组115例,以及先行急诊脑室外引流,病情稳定后再行立体定向丘脑血肿引流术的对照组108例。对比2组患者术后并发症及治疗效果的差异,评价颞部体表定位单引流管双靶点丘脑血肿脑室引流术在丘脑出血破入脑室伴脑积水治疗中的应用价值。结果研究组与对照组丘脑术后再出血率(5.2%vs.4.7%)、血肿清除率(87.5%±7.3%vs.90.2%±8.5%)、死亡率(13.9%vs.15.7%)比较差异无统计学意义(P>0.05)。研究组与对照组带管时间、术后颅内感染、分流依赖型脑积水、治疗有效、预后良好分别为(75.5±18.4)h、3.5%、19.1%、53.9%、51.3%与(130.8±22.9)h、13.9%、35.1%、38.7%、38.0%,2组比较差异有统计学意义(P<0.05)。结论 急诊颞部体表定位单引流管双靶点丘脑血肿脑室引流术治疗丘脑出血破入脑室伴脑积水,手术安全可靠,可改善患者预后。