Two-Staged Method for Ice Channel Identification Based on Image Segmentation and Corner Point Regression

Identification of the ice channel is the basic technology for developing intelligent ships in ice-covered waters,which is important to ensure the safety and economy of navigation.In the Arctic,merchant ships with low ice class often navigate in channels opened up by icebreakers.Navigation in the ice channel often depends on good maneuverability skills and abundant experience from the captain to a large extent.The ship may get stuck if steered into ice fields off the channel.Under this circumstance,it is very important to study how to identify the boundary lines of ice channels with a reliable method.In this paper,a two-staged ice channel identification method is developed based on image segmentation and corner point regression.The first stage employs the image segmentation method to extract channel regions.In the second stage,an intelligent corner regression network is proposed to extract the channel boundary lines from the channel region.A non-intelligent angle-based filtering and clustering method is proposed and compared with corner point regression network.The training and evaluation of the segmentation method and corner regression network are carried out on the synthetic and real ice channel dataset.The evaluation results show that the accuracy of the method using the corner point regression network in the second stage is achieved as high as 73.33%on the synthetic ice channel dataset and 70.66%on the real ice channel dataset,and the processing speed can reach up to 14.58frames per second.

Genesis, evolution and reservoir identification of a Neogene submarine channel in the southwestern Qiongdongnan Basin, South China Sea

A rarely reported middle-late Miocene-Pliocene channel(incised valley fill),the Huaguang Channel(HGC),has been found in the deep-water area of the southwestern Qiongdongnan Basin(QDNB).This channel is almost perpendicular to the orientation of another well-known,large,and nearly coeval submarine channel in this area.Based on the interpretation of high-resolution 3D seismic data,this study describes and analyzes the stratigraphy,tectonics,sedimentation,morphology,structure and evolution of HGC by means of well-seismic synthetic calibration,one-and two-dimensional forward modeling,attribute interpretation,tectonic interpretation,and gas detection.The HGC is located on the downthrown side of an earlier activated normal fault and grew northwestward along the fault strike.The channel is part of a slope that extends from the western Huaguang Sag to the eastern Beijiao Uplift.The HGC underwent four developmental stages:the(1)incubation(late Sanya Formation,20.4–15.5 Ma),(2)embryonic(Meishan Formation,15.5–10.5 Ma),(3)peak(Huangliu Formation,10.5–5.5 Ma)and(4)decline(Yinggehai Formation,5.5–1.9 Ma)stages.The channel sandstones have a provenance from the southern Yongle Uplift and filled the channel via multistage vertical amalgamation and lateral migration.The channel extended 42.5 km in an approximately straight pattern in the peak stage.At 10.5 Ma,sea level fell relative to its lowest level,and three oblique progradation turbidite sand bodies filled the channel from south to north.A channel sandstone isopach map demonstrated a narrow distribution in the early stages and a fan-shaped distribution in the late stage.The formation and evolution of the HGC were controlled mainly by background tectonics,fault strike,relative sea level change,and mass supply from the Yongle Uplift.The HGC sandstone reservoir is near the Huaguangjiao Sag,where hydrocarbons were generated.Channel-bounding faults and underlying faults link the source rock with the reservoir.A regionally extensive mudstone caprock overlies the channel sandstone.Two traps likely containing gas were recognized in a structural high upstream of the channel from seismic attenuation anomalies.The HGC will likely become an important oil and gas accumulation setting in the QDNB deep-water area.

Blind Channel Identification for Cyclic-Prefixed MIMO-OFDM Systems with Virtual Carriers

This paper applies the repetition index scheme(RIS)to the channel identification of cyclic prefixed(CP)multiple-input multiple-output orthogonal frequency division multiplexing(MIMO-OFDM)systems with virtual carriers(VCs)in the environment of the number of receive antennas being no less than that of transmit antennas.The VCs will cause a rank deficiency problem in computing the subspace information.With the subcarrier mapping matrix,the received signal is simplified to remove the rank deficiency.We use the RIS scheme to generate many times of equivalent symbols so the channel identification can converge with few received OFDM blocks.The RIS scheme will convert the white noise into non-white noise.With the Cholesky factorization,a noise whitening technique is developed to turn the non-white noise back to white noise.We further analyze the necessary conditions of identifiability of channel estimation.Simulations are performed to show the superiority of the proposed method.

SUBSPACE METHOD FOR BLIND IDENTIFICATION OF CDMA TIME-VARYING CHANNELS

A new blind method is proposed for identification of CDMA Time-Varying (TV)channels in this paper. By representing the TV channel's impulse responses in the delay-Doppler spread domain, the discrete-time canonical model of CDMA-TV systems is developed and a subspace method to identify blindly the Time-Invariant (TI) coordinates is proposed. Unlike existing basis expansion methods, this new algorithm does not require .estimation of the base frequencies, neither need the assumption of linearly varying delays across symbols. The algorithm offers definite explanation of the expansion coordinates. Simulation demonstrates the effectiveness of the algorithm.

System identification of channel roughness for Mid-Route Project of South-to-North Water Diversion

This paper presents a new method for the system identification of the channel roughness for the water diversion projects. According to the principle of hydraulics,the function relationship among channel roughness n, roughness height k s and hydraulic radius R is established,and then a linear model is deduced by means of the mathematical transformation to make use of the least square method for identification. Finally,based on the prototype observation data from the South-to-North Water Diversion Project and considering the influence of channel lengths,cross-section shapes and bottom slopes,etc,a universal formula is obtained for calculation of channel roughness by the system identification.

Quantitative criteria for identifying main flow channels in complex porous media

To identify the type of main flow channels of complex porous media in oil and gas reservoirs,the"main flow channel index"is defined as the ratio of comprehensive permeability obtained from well test to matrix permeability obtained from core analysis or well logging.Meanwhile,a mathematical model is established based on equivalent flow assumption,the classification method for main flow channels is put forward,and quantitative characterization of main flow channels is realized.The method has been verified by analysis of typical gas reservoirs.The study results show that the"main flow channel index"can quantitatively classify types of flow channels.If the index is less than 3,the matrix pore is the main flow channel;if the index is between 3 and 20,the fracture is the main flow channel and the matrix pore acts as the supplement one;if the index is more than 20,the fracture is the only seepage channel.The dynamic analysis of typical gas reservoirs shows that the"main flow channel index"can be used to identify the type of flow channel in complex porous media,guiding the classified development of gas reservoirs,and avoiding development risk.

A Material Identification Approach Based on Wi-Fi Signal

Material identification is a technology that can help to identify the type of target material.Existing approaches depend on expensive instruments,complicated pre-treatments and professional users.It is difficult to find a substantial yet effective material identification method to meet the daily use demands.In this paper,we introduce a Wi-Fi-signal based material identification approach by measuring the amplitude ratio and phase difference as the key features in the material classifier,which can significantly reduce the cost and guarantee a high level accuracy.In practical measurement of WiFi based material identification,these two features are commonly interrupted by the software/hardware noise of the channel state information(CSI).To eliminate the inherent noise of CSI,we design a denoising method based on the antenna array of the commercial off-the-shelf(COTS)Wi-Fi device.After that,the amplitude ratios and phase differences can be more stably utilized to classify the materials.We implement our system and evaluate its ability to identify materials in indoor environment.The result shows that our system can identify 10 commonly seen liquids with an average accuracy of 98.8%.It can also identify similar liquids with an overall accuracy higher than 95%,such as various concentrations of salt water.

Analysis of Channeling-Path Phenomena in a Complex Fault-Block Reservoir with Low Recovery Factor and High Water-Cut Ratio

Current methods for the analysis of channeling-path phenomena in reservoirs cannot account for the influence of time and space on the actual seepage behavior.In the present study,this problem is addressed considering actual production data and dynamic characteristic parameters quantitatively determined in the near wellbore area by fitting the water-cut curve of the well.Starting from the dynamic relationship between injection and production data,the average permeability is determined and used to obtain a real-time quantitative characterization of the seepage behavior of the channeling-path in the far wellbore area.For the considered case study(Jidong oilfield),it is found that the seepage capacity of the channeling-path in the far wellbore area is far less(10 times smaller)than that of the channeling-path in the near wellbore area.The present study and the proposed model(combining near wellbore area and far wellbore area real-time data)have been implemented to support the definition of relevant adjustment measures to ultimately improve oil recovery.

基于改进MobileNet的轻量级外来入侵植物识别模型研究

外来入侵植物防治的首要任务是准确识别入侵植物种类,然而外来入侵植物种类繁多,存在类间同质和类内异质现象,给技术人员甄别与防治外来入侵植物带来了挑战。为了准确、实时和高效地识别外来入侵植物,本文提出基于改进MobileNet的轻量级外来入侵植物图像识别模型(MobileNet-LW)。以专业人员鉴定得到的113种11 628幅外来入侵植物图像作为研究对象,并按照6∶2∶2的比例划分训练集、验证集和测试集。通过Retinex、旋转和高斯噪声等方法对图像进行数据增强。为了减少类间同质现象对模型误检的影响,在模型MobileNet基础上添加了SE通道注意力机制和深度连接注意力网络,提高网络对关键特征的提取能力。为了降低模型计算消耗和内存消耗,采用通道剪枝方法对网络瘦身;为了弥补剪枝后造成模型准确率降低,采用教师网络-助教网络-学生网络的形式对剪枝后的网络进行知识蒸馏,学生网络通过软知识的学习来提高识别外来入侵植物的准确率。通过消融试验测试模型的性能,利用平均准确率、平均召回率和平均F1值3个评价指标,对现阶段经典模型与改进后模型MobileNet-LW所获得的识别结果进行评价。消融试验结果显示,在相同数据集条件下,所有改进点对模型的性能都有所提升,且改进后算法在外来入侵植物图像识别中准确率提高了5.4百分点,模型参数量减少了约53%;模型对比试验表明,EfficentNet、DBTNet、ResNet-101、ConvNext和MobileNet-LW 5个模型平均准确率分别为72.3%、74.9%、76.1%、79.7%和86.1%,表明改进后的网络提高了外来入侵植物的识别准确率。基于改进MobileNet的轻量级外来入侵植物识别模型对113种外来入侵植物识别具有较高的准确率,且模型具有轻量化特点。

基于深层神经网络的信道编码类型盲识别

为了解决当前识别算法只能识别一种或者两种码字类型以及人工提取特征复杂的问题,提出了两种基于深层神经网络模型的信道编码类型识别器,即卷积神经网络(convolutional neural network,CNN)识别器和递归CNN(recursive CNN,RCNN)识别器,用于识别接收数据中不同类型的信道码字。将待识别的软解调序列作为自然语言处理中文本分类问题的句子向量进行处理,输入到预先训练好的深层神经网络识别器中进行识别,并分析了字长度对识别准确率的影响,得出了最合适的字长度。实验结果表明,两种识别器都能够有效识别接收数据中多种类型的信道编码,且在信噪比为3 dB时CNN识别器的识别准确率能够达到99%以上,而RCNN识别器在1 dB时就能够达到99%以上的识别准确率。

结合多通道MTF和CNN的框架结构损伤识别方法

为提高复杂框架结构损伤识别的准确率,提出了一种基于多通道马尔可夫变迁场(multi-channel Markov transition field,简称MCMTF)和卷积神经网络(convolutional neural network,简称CNN)的框架结构损伤识别方法。首先,采用MCMTF理论将原始一维振动信号转换为二维图像,实现数据升维和多通道数据融合;其次,以MCMTF转换后的图像数据集作为输入训练CNN模型;最后,经调参优化自动提取损伤敏感特征,并实现损伤识别。将该方法应用于IASC-ASCE Benchmark框架结构数值模型及3层钢框架结构模型试验,对比研究了多通道MTF、单通道MTF和原始数据矩阵3种数据输入方式,CNN、长短时记忆(long short term memory,简称LSTM)神经网络和深度神经网络(deep neural network,简称DNN)3种网络模型,以及噪声对框架结构损伤识别准确率的影响。结果表明:MCMTF与CNN结合方法的损伤识别准确率最优且具有良好的鲁棒性,其对Benchmark框架数值模型模拟损伤的识别准确率可达94.4%,对3层钢框架试验模型实际损伤的识别准确率可达98.4%。

基于联网收费数据的湖南省高速公路物流通道供需适应性研究

公路运输是现代物流体系中的重要组成部分,高速公路是公路运输中最重要的载体,研究高速公路货运特征,分析物流通道供给和物流需求之间的适应性,有利于推进物流行业持续向好发展,推动交通运输与现代物流深度融合。基于高速公路联网收费数据,可有效识别高速公路物流通道,并通过流量饱和度指标来评估物流通道的供需适应性。首先将高速公路货车联网收费数据基于路径识别点处打断,降低货车运行轨迹还原的计算复杂度;其次利用广度优先搜索算法、朴素贝叶斯分类理论精准还原货车运行轨迹,然后根据货车运行轨迹,计算高速公路路段货运车流量、货运量、饱和度,从物流的角度分析研究交通设施供给的适应性。以湖南省为研究对象的实例表明,湖南省高速公路物流主通道成“三纵三横”的格局分布,京港澳高速公路等路段拥堵日益严重,许广高速公路与长沙绕城高速、江干高速与浏洪高速等衔接仍有不畅。基于联网收费数据的研究方法,可以高效精准地分析与物流强相关的高速公路供需适应性情况,为物流通道规划及交通基础设施布局的理论研究和实践应用提供参考借鉴。

手机欠费用户自助缴费方案设计与应用

为避免手机欠费停机给用户带来不便,提升用户体验,减少用户离网风险,同时降低实体营业厅的运营压力和成本,湖南电信针对如何构建手机欠费用户自助缴费的绿色通道(简称手机绿通)进行了研究,设计了一套创新方案,在无须增设新设备的前提下,解决了4G网络下绿通流程中多个核心问题,并通过不断的迭代优化证明该业务系统取得了显著的成效,实现了服务资源的合理配置及客户价值与服务成本的匹配,为企业树立了良好的口碑,大大提升了企业的竞争力,被集团采纳在全国推广。

一种结构化双注意力混合通道增强的跨模态行人重识别方法

在目前跨模态行人重识别技术的研究中,大部分现有的方法会通过单模态原始可见光图像或者对抗生成图像的局部共享特征来降低跨模态差异,导致在红外图像判别中由于底层特征信息丢失而缺乏稳定的识别准确率。为了解决该问题,该文提出一种结构化双注意力可交换混合随机通道增强的特征融合跨模态行人重识别方法,利用通道增强后的可视图像作为第三模态,通过图像通道可交换随机混合增强(I-CSA)模块对可见光图像进行单通道和三通道随机混合增强抽取,从而突出行人的姿态结构细节,在学习中减少模态间差异。结构化联合注意力特征融合(SAFF)模块在注重模态间行人姿态结构关系的前提下,为跨模态表征学习提供更丰富的监督,增强了模态变化中共享特征的鲁棒性。在SYSU-MM01数据集全搜索模式单摄设置下Rank-1和mAP分别达到71.2%和68.1%,优于同类前沿方法。

基于岭回归改进规范变量分析的微震事件实时判识

微震事件判识是煤矿微震监测的基础。现有的微震监测技术大多基于单物理量变化规律而开发,在处理含有大量噪声和干扰信号的煤矿微震数据时易产生误判情况。针对该问题,基于岭回归算法改进规范变量分析(CVA)的损失函数,实现稀疏化建模,以提升模型泛化能力。采用岭回归改进CVA对多通道煤矿微震监测数据进行融合分析,进而实时判识复杂微震监测数据状态。采用模拟数据和实际煤矿微震监测数据对岭回归改进CVA进行实验验证。在基于模拟数据的实验中,随着噪声方差由5×10^(−6)增大至5×10^(−2),岭回归改进CVA的判识准确率较CVA提升了0.6%~5.4%,误报率和漏报率之和较CVA下降4.8%~17.3%。在基于实际微震监测数据的实验中,岭回归改进CVA对20个通道的微震监测数据融合分析结果能够反映出微震信号处于波动状态,验证了该方法具备微震事件实时判识能力,平均判识准确率为97.14%,较CVA高2.39%,误报率与漏报率之和为31.06%,较CVA降低0.07%,错误率为2.86%,较CVA降低2.4%。

基于AI芯片的输电通道外破隐患边缘智能监测装置研制及应用

塔吊、挖掘机等外破隐患导致输电通道事故频繁发生,有效地检测输电通道周围的外破隐患对保障输电线路安全稳定的运行意义重大。该文以边缘智能芯片为基础,研制了一种输电通道外破隐患边缘智能检测装置,并提出了一种适用于计算资源有限前端装置的轻量化外破隐患识别方法。利用深度残差网络对输电通道图像进行视觉特征提取;利用候选区域生产网络RPN捕获外破隐患目标的候选区域,再用全卷积神经网络FCN进行外破隐患的目标分类与定位。以实际采集的输电通道图像构建成样本集,进行模型测试与验证,结果表明所提方法在边缘装置中表现出良好的适用性。

低空智联网中基于多质心OpenMax的无人机开集识别方法

随着网络化、智能化的发展,无人机(Unmanned aerial vehicles, UAVs)逐渐成为低空智联网(Low-altitude intelligent network, LAIN)的重要组成部分,但如何对低空智联网中的无人机平台进行有效的管理仍面临严峻挑战。基于无人机信号中的细微特征可对无人机进行个体识别,并检测是否为非法无人机,从而实现低空智联网中无人机的身份识别和管理。针对低空领域信道环境复杂且无法提前获取非法无人机信号样本的问题,本文提出了基于差值时频和多质心OpenMax的无人机开集识别方法。首先,提出了与信道无关的差值时频特征来降低多径信道环境对射频指纹(Radio frequency fingerprinting, RFF)特征的影响,并利用数据增强提高了识别模型的准确率和鲁棒性。其次,利用多质心OpenMax替代神经网络Softmax层,以实现无人机个体的开集识别。最后,对神经网络的损失函数进行了改进,提高了开集识别准确率。本文利用真实环境采集的数据对所提算法进行了验证,在多径信道环境中开放度为0.087时,开集识别准确率达到了93.23%,与基准算法相比,准确率分别提高了7.61%和13.4%。本文提出的算法可在复杂信道环境中有效识别无人机个体并检测出首次出现的非法无人机。

利用改进RCNN卷积核的复合绝缘子缺陷识别方法

变电站复合绝缘子缺陷的检测仍然依赖于运行人员的巡检,巡检工作量大,易因视觉疲劳导致漏检。为减少计算资源消耗和缩短训练时间,通过重新组织卷积核改进了区域卷积神经网络(RCNN),提出了一种针对绝缘子裂纹形状特征的检测方法。该方法满足在训练样本数据不足的前提下,也能得到较好的卷积神经网络(CNN)训练效果,最终实现准确的裂纹识别。训练阶段采用RGB三通道分解方法,扩充训练数据集;利用中值滤波方法去除噪声;采用改进后的卷积核训练CNN。试验阶段将图片进行RGB三通道分解,并输入CNN得到确切的裂纹中心坐标、长度;采用非极大值抑制(NMS)算法去重,得到最终的裂纹识别结果。通过实例分析表明,在训练样本不足前提下,所提方法依然能达到较好的识别准确率,并能准确识别出裂纹的具体位置。

基于GIS的海宁市输电线路及通道管理系统

针对众多的输电线路,如何快速了解线路的路径走向、电压等级等台账信息;如何快速识别输电线路通道内的鸟巢、树障、人工建筑物等危险点信息;如何把输电线路信息高效、安全共享给线路设计等部门;如何快速更新输电线路通道状态信息。针对上述问题,利用GIS技术、信息技术和无人机、地物智能判别等融合手段,构建基于GIS的海宁市输电线路通道管理系统,通过此系统可极大提高输电线路管理效率和实现通道内危险物的自动识别、预警,以保证通道安全。

龙门吊轨道交叉下连续箱梁移位工序AI识别模型

为提升工程连续箱梁移位工作质量,保障工程有序顺利进行,设计龙门吊轨道交叉下的跨海通道工程连续箱梁移位工序AI识别模型。分析连续箱梁移位工序流程,将连续箱梁移位工作划分为钢筋加工、模板安装、混凝土蒸养等五个工序。利用人工智能技术中的门限玻尔兹曼机与支持向量机构建连续箱梁移位工序AI识别模型,采集连续箱梁移位工作图像,将所采集图像作为模型输入,利用卷积神经网络中的卷积层与池化层优化门限玻尔兹曼机模型,利用优化后的模型采集连续箱梁移位工作图像特征;基于不同图像特征,利用支持向量机划分图像类别,完成工序识别。实验结果显示,该模型能够准确识别连续箱梁移位工序,有效提升工程施工性能与经济效益。