基于改进Cascade R-CNN的安全帽检测算法

针对安全帽检测中,目标形状、尺度变化大,易出现漏检、误检等问题,提出了一种基于改进级联基于区域的卷积神经网络(Cascade R-CNN)的安全帽检测算法。首先,对ResNet50进行改进形成D-ResNet50,利用可变形卷积仅增加少量参数就可增大感受野的特性,对特征提取网络的C2~C5卷积层进行重塑,提高网络对目标几何变换的适应能力和特征提取能力。其次,将D-ResNet50作为主干网络引入Cascade R-CNN,形成级联目标检测器,在每个阶段对正负样本重采样,抑制误检问题。再次,对递归特征金字塔进行改进,更高效地进行多尺度特征融合,并且基于反馈信息对特征进行二次处理,增强特征表达,提高网络的分类和定位能力。最后,使用Soft-非极大值抑制(Soft-NMS)进行后处理,进一步解决漏检问题。提出的方法在Hard hat workers数据集上的AP值相比检测基线提高了3.5%,与Sparse R-CNN、TridentNet、VFnet等先进算法相比分别提升了4.7%、5.9%、2.3%等。

基于改进Cascade R-CNN算法的船舶目标检测方法

为了解决实际场景下船舶目标检测精度低的问题,本文在Cascade R-CNN算法的基础上,提出一种船舶目标检测方法 Boat R-CNN。Boat R-CNN使用带自注意力机制的Swin-Transformer Tiny网络提取图像特征,使用Soft-NMS非极大值抑制方法提升候选框过滤精度,使用Smooth_L1损失函数加速模型收敛并减少梯度爆炸情况,使用CIOU边界框回归损失提高候选框回归质量,并针对船舶目标的形状特征优化锚框的长宽比,提高锚框的生成质量。实验结果表明,Boat R-CNN算法的精度相比原版Cascade R-CNN算法提高了21.8%,相比主流Faster R-CNN算法提高了30.3%,有效提升了实际场景下的船舶目标检测精度。

Pluggable multitask diffractive neural networks based on cascaded metasurfaces

Optical neural networks have significant advantages in terms of power consumption,parallelism,and high computing speed,which has intrigued extensive attention in both academic and engineering communities.It has been considered as one of the powerful tools in promoting the fields of imaging processing and object recognition.However,the existing optical system architecture cannot be reconstructed to the realization of multi-functional artificial intelligence systems simultaneously.To push the development of this issue,we propose the pluggable diffractive neural networks(P-DNN),a general paradigm resorting to the cascaded metasurfaces,which can be applied to recognize various tasks by switching internal plug-ins.As the proof-of-principle,the recognition functions of six types of handwritten digits and six types of fashions are numerical simulated and experimental demonstrated at near-infrared regimes.Encouragingly,the proposed paradigm not only improves the flexibility of the optical neural networks but paves the new route for achieving high-speed,low-power and versatile artificial intelligence systems.

基于Cascade R-CNN的乳腺钼靶肿块检测算法研究

乳腺癌生物学特性复杂,恶性程度极高,位于女性恶性肿瘤发病率首位。乳腺钼靶肿块的X射线检查是早期确诊乳腺癌的重要方式。但乳腺钼靶肿块的检测尚处于早期阶段,现有的计算机辅助检测检测精度较低。针对这一问题,论文提出了一种基于Cascade R-CNN的乳腺钼靶肿块检测方法。实验使用南佛罗里达大学的乳房X光检查数据集,将乳腺钼靶肿块分为良性和恶性两类。通过在特征网络中加入注意力模块,提取了较为丰富的乳腺钼靶肿块特征。此外,论文提出了一种新的FPN网络FA-FPN,进一步提高了乳腺钼靶肿块病灶特征的提取,解决了深层网络在下采样中特征出现稀释的问题,提高了乳腺钼靶肿块的检测准确率。经实验验证,该模型在南佛罗里达大学的乳房X光检查数据集上的mAP值达到82.9%,在AP75下表现尤为突出。该方法在乳腺钼靶肿块的检测中具有良好的性能,可以提高乳腺钼靶肿块的检测精度,并在一定程度上避免了误检和漏检。

Multi-dimensional multiplexing optical secret sharing framework with cascaded liquid crystal holograms

Secret sharing is a promising technology for information encryption by splitting the secret information into different shares.However,the traditional scheme suffers from information leakage in decryption process since the amount of available information channels is limited.Herein,we propose and demonstrate an optical secret sharing framework based on the multi-dimensional multiplexing liquid crystal(LC)holograms.The LC holograms are used as spatially separated shares to carry secret images.The polarization of the incident light and the distance between different shares are served as secret keys,which can significantly improve the information security and capacity.Besides,the decryption condition is also restricted by the applied external voltage due to the variant diffraction efficiency,which further increases the information security.In implementation,an artificial neural network(ANN)model is developed to carefully design the phase distribution of each LC hologram.With the advantage of high security,high capacity and simple configuration,our optical secret sharing framework has great potentials in optical encryption and dynamic holographic display.

基于CSS-Cascade Mask R-CNN的有遮挡多片烟叶部位识别

烟叶部位信息是进行烟叶分级的重要参考信息,准确识别烟叶部位对实现烟叶智能分级具有重要意义。在实际的烟叶智能分级应用中,为了提高分级效率,需要对多片烟叶等级进行同步识别。受现行上料方式的限制,同步识别的多片烟叶间往往存在局部遮挡的问题,给烟叶的目标检测和部位识别带来挑战。该研究提出一种基于改进Cascade Mask R-CNN,融合通道、非局部和空间注意力机制,并引入柔性极大值抑制检测框交并操作与斯库拉交并比损失函数(SIoU)的目标检测与识别模型(CSS-Cascade Mask R-CNN)。该模型对Cascade Mask R-CNN进行了三方面的改进:一是在其骨干网络Resent101上同时引入通道、非局部、空间3种注意力机制,使网络更加关注未被遮挡且部位特征明显区域的显著度;二是将Cascade Mask R-CNN中的损失函数SmoothL1Loss替换为SIoU损失函数,将预测框与真实框之间的方向差异引入到模型训练中提高模型检测精度;三是在筛选候选框时将常规的非极大抑制(non-max-suppression)替换为柔性非极大抑制,以避免删除候选框造成信息丢失。试验结果表明,利用提出的模型对有遮挡多片烟叶进行检测和部位识别,检测框平均准确率均值(bbox_mAP50)达到了80.2%,与改进前的Cascade Mask R-CNN模型相比提高了7.5个百分点。提出的模型与多个主流的目标检测模型(YOLOvX、YOLOv3、YOLOv5、Mask R-CNN、Cascade R-CNN)相比,分别高7.1、10.2、5.8、9.2、8.4个百分点,尤其是对较难区分的下部烟叶优势明显,因此研究结果可以为有遮挡多片烟叶部位的检测识别提供参考。

基于Local Cascade Ensemble方法的胎儿健康自动分类

为更好地自动评估胎儿宫内状态,提出一种基于local cascade ensemble(LCE)方法的胎儿健康状态分类模型。选用UCI数据集,使用ADASYN方法对不平衡数据集进行填充平衡,接着结合随机森林算法对数据特征进行选择,最后使用LCE方法对胎儿状态进行自动分类。实验结果表明,所提出模型使用的方法平均准确率、精确率、召回率和F1分数分别达到了0.9554、0.9054、0.9557和0.9290,对比传统的机器学习算法能得到更好的分类效果,有效降低了误判率。

Phase-locked single-mode terahertz quantum cascade lasers array

We demonstrated a scheme of phase-locked terahertz quantum cascade lasers(THz QCLs)array,with a single-mode pulse power of 108 mW at 13 K.The device utilizes a Talbot cavity to achieve phase locking among five ridge lasers with first-order buried distributed feedback(DFB)grating,resulting in nearly five times amplification of the single-mode power.Due to the optimum length of Talbot cavity depends on wavelength,the combination of Talbot cavity with the DFB grating leads to better power amplification than the combination with multimode Fabry-Perot(F-P)cavities.The Talbot cavity facet reflects light back to the ridge array direction and achieves self-imaging in the array,enabling phase-locked operation of ridges.We set the spacing between adjacent elements to be 220μm,much larger than the free-space wavelength,ensuring the operation of the fundamental supermode throughout the laser's dynamic range and obtaining a high-brightness far-field distribution.This scheme provides a new approach for enhancing the single-mode power of THz QCLs.

Enhanced picosecond terahertz wave generation based on cascade effects in a terahertz parametric generator

Enhanced terahertz wave generation via a Stokes cascade process has been demonstrated using picosecond pulse pumped terahertz parametric generation at 1 kHz repetition rate.Clear cascade saturation of terahertz output was observed,and the corresponding cascade-Stokes spectra were analyzed.The maximum terahertz wave average power was 22μW under a pump power of 30 W,whereas the maximum power conversion efficiency was 8×10^(-7)under a pump power of 21 W.The THz power fluctuation was measured to be about 1%in 20 min.This THz parametric source with a relatively stable output is suitable for a variety of practical applications.

Conjugated microporous polymers-scaffolded enzyme cascade systems with enhanced catalytic activity

Enhancing catalytic activity of multi-enzyme in vitro through substrate channeling effect is promis-ing yet challenging.Herein,conjugated microporous polymers(CMPs)-scaffolded integrated en-zyme cascade systems(I-ECSs)are constructed through co-entrapping glucose oxidase(GOx)and horseradish peroxidase(HRP),in which hydrogen peroxide(H_(2)O_(2))is the intermediate product.The interplay of low-resistance mass transfer pathway and appropriate pore wall-H_(2)O_(2)interactions facilitates the directed transfer of H_(2)O_(2),resulting in 2.4-fold and 5.0-fold elevation in catalytic activ-ity compared to free ECSs and separated ECSs,respectively.The substrate channeling effect could be regulated by altering the mass ratio of GOx to HRP.Besides,I-ECSs demonstrate excellent stabili-ties in harsh environments and multiple recycling.

Enhancing the Electrocatalytic Oxidation of 5-Hydroxymethylfurfural Through Cascade Structure Tuning for Highly Stable Biomass Upgrading

Electrocatalytic 5-hydroxymethylfurfural oxidation reaction(HMFOR)provides a promising strategy to convert biomass derivative to highvalue-added chemicals.Herein,a cascade strategy is proposed to construct Pd-NiCo_(2)O_(4)electrocatalyst by Pd loading on Ni-doped Co3O4 and for highly active and stable synergistic HMF oxidation.An elevated current density of 800 mA cm^(-2)can be achieved at 1.5 V,and both Faradaic efficiency and yield of 2,5-furandicarboxylic acid remained close to 100%over 10 consecutive electrolysis.Experimental and theoretical results unveil that the introduction of Pd atoms can modulate the local electronic structure of Ni/Co,which not only balances the competitive adsorption of HMF and OH-species,but also promote the active Ni^(3+)species formation,inducing high indirect oxidation activity.We have also discovered that Ni incorporation facilitates the Co2+pre-oxidation and electrophilic OH*generation to contribute direct oxidation process.This work provides a new approach to design advanced electrocatalyst for biomass upgrading.

Optical facet coatings for high-performance LWIR quantum cascade lasers atλ∼8.5μm

We report on the performance improvement of long-wave infrared quantum cascade lasers(LWIR QCLs)by studying and optimizing the anti-reflection(AR)optical facet coating.Compared to the Al2O3 AR coat⁃ing,the Y_(2)O_(3)AR coating exhibits higher catastrophic optical mirror damage(COMD)level,and the optical facet coatings of both material systems have no beam steering effect.A 3-mm-long,9.5-μm-wide buried-heterostruc⁃ture(BH)LWIR QCL ofλ~8.5μm with Y_(2)O_(3)metallic high-reflection(HR)and AR of~0.2%reflectivity coating demonstrates a maximum pulsed peak power of 2.19 W at 298 K,which is 149%higher than that of the uncoated device.For continuous-wave(CW)operation,by optimizing the reflectivity of the Y_(2)O_(3)AR coating,the maximum output power reaches 0.73 W,which is 91%higher than that of the uncoated device.

基于Open CASCADE的拱桥建模研究

几何引擎是三维BIM建模软件中的核心组件,对软件的性能、准确性、互操作性和用户体验等方面有着重要影响。商业软件的几何引擎虽然经过了多年的发展和优化具有较好的性能和稳定性,但常受限于软件厂商提供的API接口和商业软件所需要的系统环境,无法对底层数据进行访问或修改,灵活性较差,且不具备完全自主商业版权。基于开源几何引擎的应用在很多领域已有相关的研究成果,但在桥梁工程领域较少,基于开源几何引擎Open CASCADE对桥梁工程领域的模型构建技术进行探索,实现自主可控的桥梁工程三维建模平台,以拱桥为研究对象,结合几何引擎的模型构建技术,实现拱桥模型的创建。探索开源几何引擎在桥梁工程领域的建模实现能力,为实现更多桥型的三维建模以及基于模型更进一步的衍生应用积累经验,并为打造自主可控的桥梁工程建模软件提供参考。

Coordinated Capacitor Voltage Balancing Method for Cascaded H-bridge Inverter with Supercapacitor and DC-DC Stage

Cascaded H-bridge inverter(CHBI) with supercapacitors(SCs) and dc-dc stage shows significant promise for medium to high voltage energy storage applications. This paper investigates the voltage balance of capacitors within the CHBI, including both the dc-link capacitors and SCs. Balance control over the dc-link capacitor voltages is realized by the dcdc stage in each submodule(SM), while a hybrid modulation strategy(HMS) is implemented in the H-bridge to balance the SC voltages among the SMs. Meanwhile, the dc-link voltage fluctuations are analyzed under the HMS. A virtual voltage variable is introduced to coordinate the balancing of dc-link capacitor voltages and SC voltages. Compared to the balancing method that solely considers the SC voltages, the presented method reduces the dc-link voltage fluctuations without affecting the voltage balance of SCs. Finally, both simulation and experimental results verify the effectiveness of the presented method.

Model Predictive Control for Cascaded H-Bridge PV Inverter with Capacitor Voltage Balance

We designed an improved direct-current capacitor voltage balancing control model predictive control(MPC)for single-phase cascaded H-bridge multilevel photovoltaic(PV)inverters.Compared with conventional voltage balanc-ing control methods,the method proposed could make the PV strings of each submodule operate at their maximum power point by independent capacitor voltage control.Besides,the predicted and reference value of the grid-connected current was obtained according to the maximum power output of the maximum power point tracking.A cost function was con-structed to achieve the high-precision grid-connected control of the CHB inverter.Finally,the effectiveness of the proposed control method was verified through a semi-physical simulation platform with three submodules.

改进的Cascade R-CNN算法在目标检测上的应用

针对Cascade R-CNN目标检测算法中存在检测精度较低以及目标遮挡问题,本文提出一种改进的Cas-cade R-CNN网络目标检测算法.该算法在主干网络ResNet101中引入可切换空洞卷积模块(Switchable Atrous Convolu-tion,SAC),该模块主要由两个全局上下文模块以及SAC组件构成,采用SAC组件以不同的空洞卷积率对特征进行卷积,并使用Switch函数收集特征来提高特征提取能力.同时,在ResNet101残差网络中引入坐标注意力机制(Coordi-nate Attention,CA),该机制将位置信息嵌入通道注意力中,用于更好地获取方向感知和位置感知信息,进而提高目标检测精度.此外,针对目标遮挡问题,引入Repulsion Loss损失函数.该损失函数主要由吸引项和排斥项组成,吸引项使得预测框和匹配上的目标框尽可能接近,排斥项使得预测框远离错误目标,进而减少非极大值抑制(Non-Maximum Suppression,NMS)的误检,提高目标检测中遮挡问题的检测精度.实验结果表明,在公开的科大讯飞AI挑战赛数据集上,与原算法测试性能相比,改进的Cascade R-CNN网络对该数据集检出率增长了2.39%,改进算法的识别精度有一定的提高.

基于改进Cascade RCNN的输电线路防振锤脱落检测方法

无人机巡检输电线路时,因拍摄角度和距离问题,容易出现被输电线遮挡和远距离拍摄的防振锤脱落目标,导致目标特征被遮挡且分辨率较低,且部分防振锤出现滑移现象,导致目标识别准确率降低。针对以上问题,提出一种基于改进Cascade RCNN的防振锤脱落检测网络。第一,设计了对比学习网络,将正负样本与真实样本的特征进行对比学习,利用对比损失函数训练网络,使其能更加关注到被遮挡的防振锤脱落目标,提升其特征提取能力;第二,进行了分类器增强操作,筛选出网络级联结构中回归效果较好的感兴趣区域并送入最后的分类回归队列中,提高了分类器的分类能力,进而提升检测目标的分类分数;第三,设计了并行注意力机制模块,整合网络提取的特征,增大关键特征的权重,使网络关注到图像中更关键的区域;在特征金字塔中,将双线性插值方法代替为反卷积,提升特征还原能力。经交叉验证实验结果表明,改进后的模型召回率、精确率和平均精度达到了97.5%,91.0%和92.0%,相比基线模型分别提高了6.9%,28.4%和8.0%。

基于改进Cascade R-CNN的绝缘子故障检测方法研究

针对航拍图中存在的绝缘子故障位置在图像中占比小、背景环境复杂导致的故障检测准确率低的问题,提出一种基于改进Cascade R-CNN模型的绝缘子故障检测方法。在原有Cascade R-CNN模型的基础上,在骨干网络中引入可变形卷积学习几何变换能力,在检测器中引入平衡损失函数平衡难易样本。在模型训练阶段,使用Copy-Paste与Mosica丰富故障绝缘子样本,平衡正负样本。使用该模型对航拍绝缘子图片进行故障检测实验,改进损失函数的模型与传统Cascade R-CNN模型相比平均召回率提升0.38%,引入可变卷积后的Cascade R-CNN模型与Faster R-CNN模型,相比平均召回率,从原来的89.78%变成93.49%,结果表明该模型能够有效克服样本遮挡以及样本不平衡的干扰。

基于改进Cascade R-CNN的探地雷达管线目标检测

针对人工识别探地雷达管线图像时效率低、误差大和成本高昂等问题,本文提出了一种基于改进Cascade R-CNN的管线目标智能化检测方法.首先对探地雷达管线图像数据集进行预处理,提升数据质量.然后采用ResNeXt代替ResNet作为主干网络提取目标特征信息,并添加多尺度特征融合模块FPN使高层特征向低层特征融合,增强低层特征表达能力.其次,使用高斯形式的非极大值抑制方法Soft-NMS得到更加精准的候选框,使用Smooth_L1作为损失函数,加速了模型收敛并且降低了训练中发生梯度爆炸的概率.最后,对于管线目标特殊的形状特征,设置合适的锚框长宽比和大小,提高锚框的生成质量.实验结果表明,本文方法解决了复杂特征的地下管线目标智能化检测,对地下管线目标检测的平均精度达到94.7%,比Cascade R-CNN方法提高了10.1%.

基于改进Cascade R-CNN的布匹瑕疵检测算法

布匹瑕疵的自动化检测是目前纺织行业面临的一个难点问题。针对当前布匹瑕疵检测算法对尺度和长宽比变化大、小目标众多的样本检测效果并不理想的问题,提出了基于改进Cascade R-CNN网络的布匹瑕疵检测算法。首先,在特征提取网络ResNet-50中融入可变形卷积,自适应地提取更多的瑕疵形状与尺度特征;其次,在特征金字塔网络上采样前引入平衡特征金字塔,缩小特征融合前各特征层之间的语义差距,得到更具表达力的多尺度特征;然后,根据瑕疵尺度与长宽比特点重新设计更适合的初始锚框;最后,采用具有尺度不变性的GIoU Loss作为级联检测器的回归损失,以获取更加精确的瑕疵预测边界框。实验结果表明,相比基于Cascade R-CNN的算法,改进后的Cascade R-CNN算法对布匹瑕疵检测的平均精确率获得了明显提升。