基于视觉SLAM和目标检测的语义地图构建

语义信息对于移动机器人理解环境内容、执行复杂任务至关重要,针对ORB-SLAM2构建的点云过于稀疏、缺乏语义信息、点云所占存储空间大等问题,提出将目标检测算法与视觉SLAM（同时定位与地图构建）技术紧密结合,构建环境的稠密点云语义地图。首先,通过目标检测网络YOLO v3及对象正则化准确获取物体的2D标签,并经过ORB-SLAM2算法构建环境的稀疏点云地图,通过含有2D标签的彩色图像和对应的深度图像以及关键帧来生成含有语义信息的稠密点云标签,使用基于图的分割算法对稠密点云进行分割,再将点云标签与分割后的点云进行融合,进而构建环境的稠密点云语义地图。文中方法在TUM公开数据集上进行试验,实验表明可以构建出效果较好的语义地图。与传统的ORB-SLAM2相比,此系统在构建地图的过程中,相机的绝对位姿误差和绝对轨迹误差分别减少了16.02%和15.86%,提高了建图精度。为了减小点云地图的存储空间,方便移动机器人进行避障和导航,最终将所构建的语义地图转换为八叉树地图。     

基于改进的局部表面凸性算法三维点云分割

点云分割是点云分类、识别以及三维重建等处理的基础,分割结果对后续应用影响巨大。本文提出利用连通点集改进局部表面凸性算法中邻近点关系的方法,解决目前激光三维成像系统点云分割算法在处理复杂环境散乱点云时存在分割过度及分割不充分的问题,通过主顶点与周围点构成连通集,作为分割判断局部子点集,形成有效分割区域。该方法解决了常用点云分割方法无法对形状不规则物体进行有效分割的问题,提高了分割精度。算法实验结果表明,相比于最小切割算法和区域生长算法,基于连通点集的改进局部表面凸性算法对实际路面环境信息的分割效果更好,并能在一定程度上避免分割过度和分割不充分的情况,证明该方法适用于复杂环境散乱点云数据分割。     

基于三维激光扫描的物体重构建模

针对锂离子电池SOC(荷电状态)难以估算的问题,通过对电池建立等效的Thevenin电路模型,对不同时刻的SOC的模型参数进行拟合得到动态的模型参数,在Matlab中借助Simulink建立仿真模型,采用模块化结构,建立基于卡尔曼滤波算法的电池SOC估算系统;利用测得的电池电压电流,仿真系统可直接估算出实时的电池SOC,与实际的电池SOC对比,误差保持在2.5%以内,表明该方法可以有效的估计电池的SOC,对于锂离子电池在实际应用的容量估算有着重要意义。     

Planar segmentation of indoor terrestrial laser scanning point clouds via distance function from a point to a plane

Terrestrial laser scanners are frequently used in most of measurement application, particularly in documentation and restoration studies of indoor historical structures, and in acquiring facade reliefs. When compared to a photogrammetric method, terrestrial laser scanners have the ability to give three dimensional point cloud data directly in a fast and detailed way. High data density of point cloud data is a challenging factor in texture-map operations during documentation and restoration of historical artifacts with more indoor spaces. When coordinate information for terrestrial laser scanner point cloud data is documented, it is seen that there is no regular order and classification for the data. The aim of this study is to suggest the mathematical filtering algorithm for segmentation work towards separation of planar surfaces which have different depths and parallel to each other and which can be frequently encountered in the indoor spaces from the data of terrestrial laser scanner. Filtering function for segmentation used, is based on the distance of a point to the plane. This algorithm has been chosen for the advantage of the rapid and easy results for extracting 3D coordinate data in texture mapping process. The MatLAB interface has been developed for using this method and analyzing the results for application which is detected how many different surfaces exist according to the statistical deviation amount. In the application, test data with 21932 points was segmented by separating it into 16 points in total with four different planes and four corner points per plane. Surfaces with four different depths were obtained as the result of the research. Each of them included four points. These segmented surfaces consisting of four points will facilitate integrated data production by integrating vectorial terrestrial laser scanner data into raster camera data, without the need to conventional measurements that accelerate particularly documentation and modeling in the fields of historical indoor areas.     

云计算Hadoop平台的异常数据检测算法研究

近年来,随着我国互联网技术的飞速发展与大规模网络运算平台研究的深入,云平台下的数据处理已成为大规模数据的主要处理方式。但是,现有的云计算Hadoop平台在海量数据异常涌入状态下,常常出现数据逻辑错误、数据链完整性缺失、数据失效的问题,造成无法对上述异常数据进行有效检测处理,严重影响云计算Hadoop平台的数据运算准确性。针对上述问题,提出云计算Hadoop平台的异常数据检测算法研究方法。采用JNS数据采集筛查模组、算法逻辑补偿模组与动态反馈模组对现有的云端计算平台存在的问题进行针对性解决。通过仿真模拟实验证明,提出的云计算Hadoop平台的异常数据检测算法研究方法,具有异常数据识别率高,准确性高,速度快、可实施性强、稳定性好的特点。     

基于多维评价模型及改进蚁群优化算法的云计算资源调度策略

针对云计算环境下资源调度模型未充分考虑资源评价的问题,为更好适应不同节点计算性能和大规模数据环境的处理需求,提出了一种基于多维评价模型的虚拟机资源调度策略。首先,在云计算环境下建立包括网络性能在内的多维资源评价模型,在此基础上提出一种改进的蚁群优化算法实现资源调度策略;然后在云计算仿真平台CloudSim上进行实现。实验结果表明,该算法可以更好适应不同网络性能的计算环境,显著提高了资源调度的性能,同时降低了虚拟机负载均衡离差,满足了云计算环境下的虚拟机资源负载均衡需求。     

土壤水分去除算法的田间原位光谱反演棉田有机质

田间原位可见-近红外光谱(VIS-NIR)能够有效的提高土壤属性的检测效率,但由于原位土壤中水分因素的影响,土壤属性的预测精度很难达到预期。如何有效去除土壤中的水分对土壤其他属性光谱预测的影响,是利用田间原位光谱高精度预测土壤属性所面临的难题,也是土壤光谱技术由室内转向田间的突破口。该问题的有效解决,可减除土壤样品的采集与室内预处理等过程,实现土壤属性的田间原位光谱测定。以新疆南部地区阿拉尔垦区十二团棉田为研究区,采用网格采样法共采集了116个0～20 cm深度的表层土壤样品,剔除1个异常值样品,得到115个有用样品,利用SR-3500型便携式地物光谱仪采集了231个样点的田间原位光谱数据,土样经风干、研磨和过筛等处理后测定其室内光谱和有机质含量。利用Kennard-Stone算法将115个土样分为69个转换子集及46个预测集,采用外部参数正交化法(EPO)、光谱直接转换法(DS)及光谱间接转换法(PDS)三种去除水分算法结合原位光谱反射率(R)、反射率一阶微分(R′)、反射率对数(LOG(R))以及反射率倒数(1/R)四种数学变换方式,运用随机森林(RF)模型进行不同组合模型的构建及精度评价。结果表明：(1)土壤有机质含量越高,土壤光谱反射率越低。土壤田间原位光谱反射率低于土壤室内光谱反射率;(2)室内光谱反射率与土壤有机质含量之间的相关性大于田间原位光谱,室内光谱经一阶微分变换后与土壤有机质含量之间的相关性显著提升。(3)土壤室内光谱反射率模型预测精度(R2=0.86, RPD=2.08, RMSE=1.55 g·kg-1, MAPE= 0.14)高于田间原位光谱反射率模型(R2=0.71, RPD=1.49, RMSE=2.17 g·kg-1, MAPE=0.20)。在去除水分算法模型中,以EPO一阶微分模型去除水分效果最好,决定系数R2由0.71提高到0.83,RPD由1.49提高到2.04,RMSE由2.17 g·kg-1降低至1.58 g·kg-1,MAPE由0.20降低至0.14。本研究实现了去除土壤水分因素的影响,提高了田间原位光谱预测土壤有机质的精度,为南疆棉田大尺度土壤有机质的预测及土壤肥力的评价提供了重要的参考。     

基于机载激光雷达点云的雪道坡度提取算法

滑雪场的雪道坡度信息不仅是雪场规划和建设所需的必要内容,也是滑雪运动员提高成绩、减少运动损伤的重要数据。传统的手动测量方式费时费力,且雪场的低温环境不利于测量人员长时间作业。为了解决滑雪场的雪道坡度计算问题,提出了一种基于激光雷达点云的雪道坡度自动提取算法。使用投影高程差滤波、聚类分割等算法对雪场点云进行预处理,获取雪道边缘特征数据。提出了一种自适应阈值的拟合算法,用于雪道中线的提取,该算法可根据点云数量信息及坐标信息,自动选择阈值拟合雪道中线,消除了实际应用中部分点云数据稀疏或缺失对中线拟合带来的不利影响。数据处理结果表明,该方法计算所得坡度值与多次手动测量平均值之间的相对误差为2.2%。这一方法对激光雷达在雪场目标上的应用、雪场测绘中雪道坡度计算等具有参考意义。     

Multiple small objects tracking based on dynamic Bayesian networks with spatial prior

This paper proposes an end-to-end algorithm for multiple small objects tracking in noisy video using a combination of Gaussian mixture based background segmentation along with a Dynamic Bayesian Networks (DBNs) based tracking. Background segmentation is based on an adaptive backgrounding method that models each pixel as a mixture of Gaussians with spatial prior and uses an online approximation to update the model, the spatial prior is constructed for small objects. Furthermore, we create observation model with hidden variable based on multi-cue statistical object model and employ Kalman filter as inference algorithm. Finally, we use linear assignment problem (LAP) algorithm to perform the models matching. The experimental results show the proposed method outperforms competing method, and demonstrate the effectiveness of the proposed method.     

基于图像的树木测量方法

树木是城市绿化、美化的一个重要的组成部分,具有重要的科学、历史和观赏价值。传统的树木生长信息采集方式,利用人工使用卷尺或不易操作、低效率和复杂麻烦的大型测量仪器等,费时、费力、准确度又不高。结合计算机视觉、数学几何模型等技术,设计了一种新式的树木高度、胸径和冠幅测量模型,提出了基于图像的树木测量方法和实现了树木信息采集软件系统;为了得到更精确的测量结果,使用摄像头标定技术和坐标校正方法对图像进行畸变矫正;并且利用正交试验分析法分析本设计使用的最佳拍摄环境,最终相对误差控制在4%左右。本方法具有携带方便、操作简单、适用面广、准确度高的优势。     

可迁移的土壤重金属污染高光谱定性分类方法研究

现有基于高光谱遥感的土壤重金属污染定性分类模型,大多采用同一地区室内光谱测定训练样本数据进行模型构建与测试。但室内光谱测定需要复杂的处理过程,成本高,效率低,且无法快速获得目标区域空间上连续的光谱信息。考虑到实际应用需求,模型在相同实验区和不同试验区野外光谱数据是否具有较好的迁移推广能力是目前迫切需要回答的问题。为回答这一问题,选取湖南省郴州市和衡阳市两铅锌矿区作为实验研究区,选用支持向量机（SVM）作为分类器,将郴州实验区室内采样的83个样本数据和衡阳实验区室内采样的46个样本数据分别用于分类器训练,将衡阳地区野外采样的46个样本数据用于分类测试。并首先通过基于联合分布适配（JDA）的迁移学习方法进行光谱变换以缩小两地室内外测定光谱分布差异,然后进行不同区域室内外土壤重金属污染定性分类模型迁移。实验结果表明：（1）由于野外测得的光谱数据会受到太阳辐射、提取的土壤成分差异等因素的干扰导致室内外光谱数据存在显著的分布差异,难以直接将基于室内采样数据训练得到的土壤重金属污染定性分类模型迁移到同一地区测定的野外高光谱数据上。但通过JDA变换缩小室内外分布差异后,模型迁移能力得到显著提升,砷（As）、铅（Pb）和锌（Zn）三种重金属含量是否超标的分类精度都达到了84%以上,Zn元素含量是否超标的分类精度甚至达到了89%以上。（2）由于季节性影响、地区成分的干扰和光谱噪声的增加,不同地区光谱数据存在着更为显著的分布差异,加大了不同地区土壤重金属污染监测的难度,难以将基于室内采样光谱数据所建立的土壤重金属定性分类模型直接迁移到其他地区野外采样数据上（平均分类精度仅在50%左右）。经过JDA迁移学习方法进行室内外光谱变换处理后,模型迁移能力得到保证,因此,室外光谱采样可直接用于研究不同试验区域重金属（As,Pb和Zn）的污染情况。     

多源信息融合的环境舒适度检测与评价

为了能够实现对室内环境的监测和舒适度的调节,首先需要获取准确、及时的室内环境参数。针对目前室内环境监测系统存在自动化程度低、检测点部署欠灵活,安装维护困难,监控难度大等问题,设计开发了一种基于无线传感网络的室内环境监测系统。以ZigBee技术为核心,构建无线传感器网络,自动采集和传输室内相关环境参数;通过结合物联网云平台,实现远程监控和报警;运用模糊综合评价的方法,通过统计、分析、计算,得到符合实际的室内环境舒适度水平。本系统通过硬件设计降低了室内环境监测成本,通过软件设计提高了系统的监测水平和灵活性,通过结合模糊综合评价方法和物联网云平台降低了监控难度。经实验,验证了本系统在实际环境下的使用性能,适合推广应用。     

A quantification model of overall dissatisfaction with indoor noise environment in residential buildings

A quantification model concerning overall dissatisfaction from multiple noise sources in residential buildings and underlying assumptions were presented. The model was constructed by two steps; a survey and an auditory experiment. The relation between dissatisfaction with the indoor noise environment and dissatisfaction with individual noises such as floor impact, airborne, drainage, and traffic noises was first found in the survey. The annoyance from individual noises was obtained as a function of the noise level from the laboratory experiment. Then, annoyance ratings were translated into the percentage of dissatisfaction based on the relation between annoyance and dissatisfaction obtained from the survey. Finally, equations were derived for predicting the degree of dissatisfaction with the overall indoor noise environment using individual noise levels, and a classification method for the noise environment with multiple noise sources were proposed. The procedure and the quantification model can be used for the assessment of the associated overall dissatisfaction of the indoor noise environment on the basis of the level of individual sources.     

Saliency Snake: A unified framework with adaptive initial curve

Active contour model also known as Snake, is a popular approach for contour extraction and image segmentation. However, some existing active contour models are sensitive to the initial curve, which is usually required to be placed near the true object contour. To address this problem, this paper proposes a novel active contour model with adaptive initial curve, namely Saliency Snake. Taking visual saliency into consideration, prior shape information of the interested object is incorporated explicitly into the energy functional of Saliency Snake. Such improvement facilitates active contour evolution and leads to fast and automatic segmentation. Experiment results demonstrate that the proposed Saliency Snake can achieve superior segmentation performance both in terms of accuracy and efficiency.     

核燃料棒反光表面条纹自适应中心提取方法

针对水下检测中激光条纹中心难以准确提取的问题,提出了一种适用于水下核燃料棒反光表面的条纹自适应中心提取方法。根据检测环境中存在水体散射及物体表面高反光的特点,去除水下噪点、反光噪点,实现激光条纹分割提取;充分利用BP神经网络曲线拟合及由光条几何信息生成的自适应卷积模板,实现反光区域的轮廓与灰度分布修正,使光条截面灰度分布符合高斯分布;经灰度重心法在光条截面方向实现激光条纹中心的亚像素精度定位与提取。实验结果表明:该方法有效地解决了被测物体反光表面光条中心线不连续、噪点多的问题,点云三维重建误差在0.2 mm以内,保证了条纹中心提取的精度与稳定性,满足核燃料棒水下检测工程要求。     

基于反射率特性的高光谱遥感图像云检测方法研究

为了提高光谱遥感图像数据的利用率,提出了一种基于光谱反射率特性的星载高光谱图像云检测方法。首先分析了各类地物在不同光谱下的反射率特征,根据其特点从星载高光谱图像中选取特定5个波长的图像数据,然后依照图像地物反射率及阈值判定法区分出高云区和中低云区,最后对二者合并,得到目标云区。试验结果表明,该方法的云检测率可达90%以上,虚警率在1%以内,可以准确区分云与雪地区域、沙漠区域以及植被区域,实现云检测。     

A novel approach to segmentation and measurement of medical image using level set methods

The study proposes a novel approach for segmentation and visualization plus value-added surface area and volume measurements for brain medical image analysis. The proposed method contains edge detection and Bayesian based level set segmentation, surface and volume rendering, and surface area and volume measurements for 3D objects of interest (i.e., brain tumor, brain tissue, or whole brain).Two extensions based on edge detection and Bayesian level set are first used to segment 3D objects. Ray casting and a modified marching cubes algorithm are then adopted to facilitate volume and surface visualization of medical-image dataset. To provide physicians with more useful information for diagnosis, the surface area and volume of an examined 3D object are calculated by the techniques of linear algebra and surface integration. Experiment results are finally reported in terms of 3D object extraction, surface and volume rendering, and surface area and volume measurements for medical image analysis.     

基于序列图像的运动目标检测

针对动态环境下运动目标检测中噪声多、目标检测不完整等情况,提出了一种基于金字塔多分辨率模型的运动目标检测方法,在低分辨率下获取目标的区域,在高分辨率下获取目标的细节。对于复杂的环境,还提出了运用高低双阈值替代传统的单阈值进行图像差分运算的方案,阈值可以根据图像自动分析得到。该方法首先将当前帧和背景帧进行尺度变换,得到不同分辨率下的图像组,然后在不同尺度下得到高低阈值差分图像,最后从高层向低层进行有效融合,得到噪声少的完整目标图像。实验表明,该方法提取运动目标的精度比较高,单目标达到0.802,多目标达到0.615,尤其是在复杂的动态环境下,优势比较明显。     

贝叶斯概率图像自动分割研究

探讨了一种新的图像自动分割的方法。提出应用高斯有限混合模型与期望－极大化算法对图像特征空间的数据进行聚类，采用信息理论准则（ITC）确定要分割的图像区域数目，用贝叶斯概率分割图像。整合这些技术可以实现图像自动分割，而且实验结果表明信息理论准则可以确定适当的区域数目。     

A greedy-model-based reinforcement learning algorithm for Beyond-5G cooperative data collection

Data collection is an essential part of Beyond-5G and Internet of Things applications. In urban area, heterogeneous access points such as Wi-Fi routers and base stations can meet the required communication coverage and bandwidth in data collection processes. However, in remote area, without communication infrastructures, it is hard to guarantee the communication quality of a large-scale data aggregation network. An existing approach is to use an unmanned aerial vehicle (UAV) to act as a mobile sink to perform data collection and increase the coverage of intelligent wireless sensing and communications. The efficiency and the reliability of such a UAV-assisted data collection system can be significantly enhanced with an intelligent cooperative strategy for the sensors deployed in the field to communicate with the UAV. Furthermore, an energy-efficient trajectory planning algorithm is crucial to address the physical limitations of the UAV in this application. In this paper, a data collection process is modeled as a Markov decision process (MDP). The paper begins with proposing two heuristic greedy algorithms, namely distance-greedy (DG) algorithm and rate-greedy (RG) algorithm, which are designed based on prior knowledge of the system and can guarantee the completion of the data collection process in a remote area without the help of fixed communication infrastructures. Based on the outcomes, a multi-agent greedy-model-based reinforcement learning (MG-RL) algorithm is proposed, which specifically designs the environmental state and the reward scheme, and introduces multiple UAVs with different parameters to explore environments in parallel to accelerate the training. In conclusion, the two proposed greedy algorithms have lower complexity of implementation while the proposed MG-RL algorithm yields practical UAVs’ flight trajectories and shortens the time for completing a data collection task.