基于车载环境的交通目标跟踪

针对车载环境下小目标难以识别和相机动态移动造成的目标跟踪精度下降问题, 提出一种基于改进YOLOv5与ByteTrack的交通目标跟踪方法. 首先, 引入Transformer与加权特征金字塔(BiFPN)结构的思想重构YOLOv5检测网络, 有效捕获了特征的全局依赖关系, 缓解了深层卷积小目标信息丢失问题, 改善了车载环境下的目标检测性能. 此后, 以ByteTrack为基础提出了添加相机移动补偿的CMC-ByteTrack跟踪方法, 更精准地描述了视频前后帧的数据关联关系, 提高了相机大幅位移时的跟踪精度. 实验结果表明, 改进YOLOv5的平均检测精度(mAP)达到了82.2%, 相比原算法提高了3.9%, 与CMC-ByteTrack结合后的跟踪准确性(MOTA)相比改进前的跟踪方法提高了2.8%.     

A 20.8-Gbps Dual-Carrier Wireless Communication Link in 220-GHz Band

With the successful demonstration of terahertz (THz) high-speed wireless data transmission,the THz frequencies are now becoming a worth candidate for post-5G wi...     

Systematic adsorption process of petroleum hydrocarbon by immobilised petroleum-degradation bacteria system in degradation pathways

In order to effectively control petroleum hydrocarbon pollution by immobilisation technology,it is necessary to understanding the degradation pathways of petrol...     

Key Residues Affecting Binding Affinity of Sirex noctilio Fabricius Odorant-Binding Protein (SnocOBP9) to Aggregation Pheromone

Sirex noctilio Fabricius (Hymenoptera Siricidae) is a major quarantine pest responsible for substantial economic losses in the pine industry. To achieve better pest control, (Z)-3-decen-ol was identified as the male pheromone and used as a field chemical trapping agent. However, the interactions between odorant-binding proteins (OBPs) and pheromones are poorly described. In this study, SnocOBP9 had a higher binding affinity with Z3D (Ki = 1.53 ± 0.09 μM) than other chemical ligands. Molecular dynamics simulation and binding mode analysis revealed that several nonpolar residues were the main drivers for hydrophobic interactions between SnocOBP9 and Z3D. Additionally, computational alanine scanning results indicated that five amino acids (MET54, PHE57, PHE71, PHE74, LEU116) in SnocOBP9 could potentially alter the binding affinity to Z3D. Finally, we used single-site-directed mutagenesis to substitute these five residues with alanine. These results imply that the five residues play crucial roles in the SnocOBP9-Z3D complex. Our research confirmed the function of SnocOBP9, uncovered the key residues involved in SnocOBP9-Z3D interactions, and provides an inspiration to improve the effects of pheromone agent traps.     

基于支持向量机的键盘密码输入异常检测方法研究

随着电子商务的迅速发展,网络交易安全越来越收到重视,一旦用户帐号和密码被非法窃取,将蒙受巨大损失.本文介绍了一种基于支持向量机(Support Vector Machines,SVM)的方法对密码输入过程中键盘敲击频率等特征进行识别,以此判断密码输入者的身份.实验表明这种方法可以比较有效地检测异常密码输入.     

直流输电系统用直流分压器现场检测方法研究

对直流输电系统中所用直流分压器设备的准确度进行现场检测,目前国内还没有检定规程或校准规范和实施程序。根据现场环境,提出使用桥式电阻分压器自校准的方法,提出现场实施程序,注意事项,主要仪器的选型及误差分析,分析了现场存在的影响因素,以及影响程度,并完成模拟现场环境的验证试验,试验数据表明,该方法及其实施过程符合直流分压器现场检测要求,对现场检测试验具有指导意义,并可为实验室检定提供参考价值。     

微机控制太阳能逆变电源

介绍了一种微机控制太阳能逆变电源,在这里微计算机按照给定的功率来控制逆变电源的输出电流,由于采用了锁相环技术,保证了逆变电源的输出与电网电压严格同步。控制器选用 Ｉｎｔｅｌ ８９Ｃ５１单片机,并在实验室中进行了实验,实验结果证明理论分析是正确的。     

基于遗传算法和支持向量机的特征子集选择方法

在模式分类系统中,往往需要从大量的特征中选择最优的特征子集,人工选择特征的方法往往费时费力,本文采用遗传算法(GA)对支持向量机进行封装的方法选择特征子集。首先使用遗传算法随机产生若干特征子集,通过选择、交叉和变异操作产生新的特征子集,经过若干代之后,得到最优的特征子集。在遗传算法中最重要的是适应度的确定,本文用支持向量机(SVM)作为分类器,为了避免出现“过拟和”,把特征子集的5阶交叉验证分类准确率和特征数量的联合函数作为适应度函数。对UCI机器学习库中SONAR和LED数据集进行实验,结果表明本方法可以有效滤除无关特征并提高分类准确率。     

On scheduling all-to-all personalized connection and cost-effectivedesigns in WDM rings

We consider the problem of scheduling all-to-all personalized connections (AAPC) in WDM rings. Scheduling one connection for every source-destination pair in a network of limited connectivity provides a way to reduce routing control and guarantee throughput. For a given number of wavelengths K and a given number of transceivers per node T, we first determine the lower bound (LB) on the schedule length, which depends on both K and T. To achieve the LB, either the network bandwidth, the I/O capacity, or both should be fully utilized. This approach first constructs and then schedules circles, each of which is formed by up to four non-overlapping connections and can fully utilize the bandwidth of one wavelength. The proposed circle construction and scheduling algorithms can achieve the LB if K⩽T相似文献   

浮选图像中泡沫边缘分割

矿物浮选过程中泡沫的表面特征是浮选性能的重要指标,它可以实时、直观地反映浮选效果的变化,准确分割泡沫的边缘信息是浮选过程中一项重要的任务。近年来,研究人员提出了各种浮选泡沫图像分割算法,但浮选泡沫图像中存在泡沫数量多、泡沫间粘连严重以及边缘不清晰等问题,现存的方法由于其特征提取能力有限,无法精确的分割泡沫边缘。据此,本文利用深度学习提出了一种基于多尺度融合的浮选泡沫图像边缘分割算法,该算法通过引入一种深度高分辨率的编码结构以及一种基于注意力的分层融合方法来增强模型的特征提取能力,从而提高对于浮选泡沫边缘的分割效果。具体而言,深度高分辨率的编码结构可以在不同分辨率层级上同时维护特征信息,使我们的网络模型可以有效地捕捉不同尺度的信息,在提高图像语义理解能力的同时能够保持更多的细节信息,提高处理高分辨率以及密集任务图像的能力。除此之外,本文设计了一种基于注意力的分层融合方法来充分融合深层和浅层的特征图,使融合得到的特征图趋向于更重要的特征信息,从而提高识别浮选泡沫的边界和精确定位浮选泡沫的能力。该算法在泡沫边界分割数据集上凭借58.25的泡沫IoU以及73.62的泡沫Fscore取得了最佳的性能,证明了我们提出的算法可以更加准确地分割浮选泡沫边缘。