Autonomous Salient Feature Detection through Salient Cues in an HSV Color Space for Visual Indoor Simultaneous Localization and Mapping

For successful simultaneous localization and mapping (SLAM), perception of the environment is important. This paper proposes a scheme to autonomously detect visual features that can be used as natural landmarks for indoor SLAM. First, features are roughly selected from the camera image through entropy maps that measure the level of randomness of pixel information. Then, the saliency of each pixel is computed by measuring the level of similarity between the selected features and the given image. In the saliency map, it is possible to distinguish the salient features from the background. The robot estimates its pose by using the detected features and builds a grid map of the unknown environment by using a range sensor. The feature positions are stored in the grid map. Experimental results show that the feature detection method proposed in this paper can autonomously detect features in unknown environments reasonably well.     

General Feature Extraction for Mapping and Localization of a Mobile Robot Using Sparsely Sampled Sonar Data

This study introduces a method of general feature extraction for building a map and localization of a mobile robot using only sparsely sampled sonar data. Sonar data are acquired by using a general fixed-type sensor ring that frequently provides false returns on the locations of objects. We first suggest a data association filter that can classify sets of sonar data that are associated with the same hypothesized feature into one group. A feature extraction method is then introduced to decide the exact geometric parameters of the hypothesized feature in the group. We also show the possibility of extracting a circle feature consistently as well as a line or a point feature by using the proposed filter. These features are then assembled to build a global map and applied to extended Kalman filter-based localization of the robot. We demonstrate the validity of the proposed filter with the results of mapping and localization produced by real experiments.     

基于深度感知特征提取的室内场景理解

从深度图RGB-D域中联合学习RGB图像特征与3D几何信息有利于室内场景语义分割,然而传统分割方法通常需要精确的深度图作为输入,严重限制了其应用范围.提出一种新的室内场景理解网络框架,建立基于语义特征与深度特征提取网络的联合学习网络模型提取深度感知特征,通过几何信息指导的深度特征传输模块与金字塔特征融合模块将学习到的深...     

Novel Sonar Salient Feature Structure for Extended Kalman Filter-Based Simultaneous Localization and Mapping of Mobile Robots

Abstract

Not all line or point features capable of being extracted by sonar sensors from a cluttered home environment are useful for simultaneous localization and mapping (SLAM) of a mobile robot. This is due to unfavorable conditions such as environmental ambiguity and sonar measurement uncertainty. We present a novel sonar feature structure suitable for a cluttered environment and the extended Kalman filter (EKF)-based SLAM scheme. The key concept is to extract circle feature clouds on salient convex objects by sonar data association called convex saliency circling. The centroid of each circle cloud, called a sonar salient feature, is used as a natural landmark for EKF-based SLAM. By investigating the environmental inherent feature locality, cylindrical objects are augmented conveniently at the weak SLAM-able area as a natural supplementary saliency to achieve consistent SLAM performance. Experimental results demonstrate the validity and robustness of the proposed sonar salient feature structure for EKF-based SLAM.     

EO-1 Hyperion数据的预处理、特征提取和岩性填图研究

EO-1 Hyperion传感器是第一个可以获取可见光与近红外以及短波红外波长范围光谱信息的星载高光谱传感器。本文以美国最早的金矿采矿区之一,加利福尼亚州东南巧克力山的Rainbow金矿区作为研究案例,探讨了Hyperion数据的预处理方法,专题信息提取与填图,评估了Hyperion高光谱数据在识别与金矿有关的岩性类型的应用价值。结果表明,本文所提出的Hyperion数据预处理方法是有效的,MNF方法能有效用于Hyperion数据维数的降低和数据冗余的去除以及分类特征的提取。最大似然分类器能够有效地从Hyperion高光谱数据中提取与金矿相关的重要岩体信息,所得到的岩性单元与地质图上对应的岩性分布具有很好的一致性。岩体分类的总精度为86%。该研究表明,Hyperion高光谱数据能够很好识别有细微光谱差别的岩性,因而在地质学研究与找矿领域有着良好的应用前景。     

结合CSS与傅里叶描述子的手势特征提取

目前常用的基于视觉的静态手势特征提取方法只从单一方面进行描述,缺乏全局信息和局部信息的有效结合。为此,提出一种结合CSS形状描述子与傅里叶描述子的手势特征提取方法。将CSS形状描述子与傅里叶描述子相结合,以此作为一种融合手势局部特征和全局特征的新的静态手势特征。实验结果表明,与传统方法相比,该方法的正确率更高,达到98.3%。     

双通道深度主题特征提取的文章推荐模型

针对传统的文章推荐方法存在的冷启动、用户反馈稀疏以及相似度衡量准确性欠佳等问题,本文提出了融合主题模型和预训练模型BERT的文章相似度计算模型(contextualized topic BERT, ctBERT).给定查询,该算法会计算查询与相关文章之间的相似度分数,文章经过预处理分别输入独立的子模块进行特征抽取并计算相似度得分,相似度得分与支撑集的个性化得分相结合以获得最终分数,该方法将单样本学习整合进推荐框架中,进一步取得了显著的改进.本文在3个不同的数据集上的实验结果表明,所提出方法在3个数据集上的NDCG标准均有提升,例如在Aminer数据集上NDCG@3和NDCG@5标准比对比方法分别提高了6.1%和7.2%,验证了该方法的有效性.     

特征提取与模板匹配结合的图像拼接方法

本文提出一种特征点与模板匹配相结合的图像拼接方法,先对相邻两幅图像利用Harris算子提取特征点,然后根据特征点的位置确定模板的大小和位置,大大减小了图像拼接的计算量,提高了拼接速度。用两幅相邻的月球表面图像进行实验,实验结果表明本文算法能取得较好的效果。     

基于中心距离特征的联机手写维吾尔文全形态字母识别研究

为了提高联机手写维吾尔文字母的正确识别率,根据维吾尔文字母的手写特点,提出了中心距离特征CDF（Center Distance Feature）、并基于CDF进行了一系列识别实验。在实验中,该文采集了400个人的手写字母样本,利用CDF的三种不同的实施方案（CDF-2,CDF-4,CDF-8）分别对维吾尔文字的32个母独立形态和128个所有形态进行了识别实验,并对实验结果进行了分析。实验结果表明,CDF是一种非常适合于维吾尔文字母识别的特征,有待于进一步改进和优化。     

基于成对约束半监督降维的高光谱遥感影像特征提取

半监督降维(Semi\|Supervised Dimensionality Reduction,SSDR)框架下,基于成对约束提出一种半监督降维算法SCSSDR。利用成对样本进行构图,在保持局部结构的同时顾及数据的全局结构。通过最优化目标函数,使得同类样本更加紧凑\,异类样本更加离散。采用UCI数据集对算法进行定量分析,发现该方法优于PCA及传统流形学习算法,进一步的UCI数据集和高光谱数据集分类实验表明:该方法适合于进行分类目的特征提取。     

仿人形机器人火炬传递中的高精度目标特征提取

以仿人形机器人的火炬传递为背景,针对立体视觉系统的图像处理展开研究,提出了一种基于色标的高精度特征提取方法.利用火炬上的矩形色标,在RGB空间基于色彩进行图像分割,并利用K-L变换对边缘点进行分组,通过Hough变换和最小二乘法对色标边缘进行直线拟合,得到矩形色标顶点的高精度图像坐标,为提高火炬位姿测量的精度提供了基础.实验结果验证了本文方法的有效性.     

Detection of Parkinson’s Disease with Multiple Feature Extraction Models and Darknet CNN Classification

Parkinson’s disease (PD) is a neurodegenerative disease in the central nervous system. Recently, more researches have been conducted in the determination of PD prediction which is really a challenging task. Due to the disorders in the central nervous system, the syndromes like off sleep, speech disorders, olfactory and autonomic dysfunction, sensory disorder symptoms will occur. The earliest diagnosing of PD is very challenging among the doctors community. There are techniques that are available in order to predict PD using symptoms and disorder measurement. It helps to save a million lives of future by early prediction. In this article, the early diagnosing of PD using machine learning techniques with feature selection is carried out. In the first stage, the data preprocessing is used for the preparation of Parkinson’s disease data. In the second stage, MFEA is used for extracting features. In the third stage, the feature selection is performed using multiple feature input with a principal component analysis (PCA) algorithm. Finally, a Darknet Convolutional Neural Network (DNetCNN) is used to classify the PD patients. The main advantage of using PCA- DNetCNN is that, it provides the best classification in the image dataset using YOLO. In addition to that, the results of various existing methods are compared and the proposed DNetCNN proves better accuracy, performance in detecting the PD at the initial stages. DNetCNN achieves 97.5 % of accuracy in detecting PD as early. Besides, the other performance metrics are compared in the result evaluation and it is proved that the proposed model outperforms all the other existing models.     

面向胎盘植入产前诊断的医学语义特征提取算法

胎盘植入由于其临床特征隐匿,尚无一种敏感性、特异性高的产前诊断手段,因此文中将数据的特征提取方法引入胎盘植入产前诊断领域,从特征相关性的角度,提出胎盘植入有效医学语义的多目标特征优化问题,并给出求解该问题的一种改进的非支配排序遗传算法II( NSGA-II)。基于实际胎盘植入相关临床数据的计算结果表明,文中算法能从复杂的胎盘植入相关临床数据中提取具有胎盘植入有效语义的特征集合。经过接收者操作特征( ROC)曲线分析,提取的特征医学语义具有较高的诊断价值,可为产科医师研究胎盘植入的发病机制和及时产前诊断提供有效的辅助手段。文中研究还发现,一些临床生化检查指标具有重要作用,可作为胎盘植入产前诊断的有效依据。