Video analysis based on volumetric event detection

During the past decade, feature extraction and knowledge acquisition based on video analysis have been extensively researched and tested on many applications such as closed-circuit television(CCTV)data analysis, large-scale public event control, and other daily security monitoring and surveillance operations with various degrees of success. However, since the actual video process is a multi-phased one and encompasses extensive theories and techniques ranging from fundamental image processing, computational geometry and graphics, and machine vision, to advanced artificial intelligence, pattern analysis, and even cognitive science, there are still many important problems to resolve before it can be widely applied. Among them, video event identification and detection are two prominent ones. Comparing with the most popular frame-to-frame processing mode of most of today's approaches and systems, this project reorganizes video data as a 3D volume structure that provides the hybrid spatial and temporal information in a unified space. This paper reports an innovative technique to transform original video frames to 3D volume structures denoted by spatial and temporal features. It then highlights the volume array structure in a so-called "pre-suspicion" mechanism for a later process. The focus of this report is the development of an effective and efficient voxel-based segmentation technique suitable to the volumetric nature of video events and ready for deployment in 3D clustering operations. The paper is concluded with a performance evaluation of the devised technique and discussion on the future work for accelerating the pre-processing of the original video data.     

A hybrid data-fusion system using modal data and probabilistic neural network for damage detection

This paper addresses a novel hybrid data-fusion system for damage detection by integrating the data fusion technique, probabilistic neural network (PNN) models and measured modal data. The hybrid system proposed consists of three models, i.e. a feature-level fusion model, a decision-level fusion model and a single PNN classifier model without data fusion. Underlying this system is the idea that we can choose any of these models for damage detection under different circumstances, i.e. the feature-level model is preferable to other models when enormous data are made available through multi-sensors, whereas the confidence level for each of multi-sensors must be determined (as a prerequisite) before the adoption of the decision-level model, and lastly, the single model is applicable only when data collected is somehow limited as in the cases when few sensors have been installed or are known to be functioning properly. The hybrid system is suitable for damage detection and identification of a complex structure, especially when a huge volume of measured data, often with uncertainties, are involved, such as the data available from a large-scale structural health monitoring system. The numerical simulations conducted by applying the proposed system to detect both single- and multi-damage patterns of a 7-storey steel frame show that the hybrid data-fusion system cannot only reliably identify damage with different noise levels, but also have excellent anti-noise capability and robustness.     

SIFT在行星表面探测器视觉导航系统中的应用

介绍了行星探测器自主导航系统的相关背景,讨论了行星探测器自主视觉导航系统算法的主要目的和传感器信息融合的必要性,并就此引入了尺度不变特征变换(SIFT)的相关理论分析.对SIFT方法提出了改进,从而改善了算法的效果.最后给出了三维重建和目标追踪的实验结果.     

基于角点检测和自适应阈值的新闻字幕检测

目前用于提取新闻视频帧中字幕的方法准确率和检测速度普遍较低,尤其对于分辨率和对比度较小的标题文字,检测效果很差。针对上述问题,提出一种基于角点检测和自适应阈值的字幕检测方法。该方法利用角点检测确定标题帧中的文字区域并进行灰度变换,利用自适应阈值的方法对其进行二值化,得到OCR可识别的文字图片。实验表明,该方法可以快速有效地提取出分辨率和对比度较小的新闻视频标题字幕。     

基于参数重要度的多元时间序列相似性查询

针对多元时间序列的相似性查询问题,给出参数重要度的定义,提出一种基于参数重要度的候选集查询方法。通过对多元时间序列的SVD分解,将奇异值向量和特征矩阵作为多元序列的特征,基于线性空间中的坐标变换原理构造2个多元时间序列的相似性度量模型,实现在候选集上的精确匹配并获得最终的结果集。对飞行数据的相似性查询实验验证了该方法的有效性。     

改进的形态学骨架提取算法

针对由传统最大圆盘骨架提取算法提取出的骨架连通性差且无法保持一致的单像素宽度问题,提出一种改进的形态学骨架提取算法。将连通性保持与形态学运算相结合,在收缩目标提取骨架的过程中通过引入虚拟骨架点实现骨架曲线连通性保持,而单像素宽度细化及伪分支剔除等后处理过程的引入则进一步提高骨架描述目标形状及拓扑特征的能力。相关仿真研究证明了该算法的有效性。     

基于统计不相关矢量集的独立成分分析

针对随机数据在高斯分布条件下,独立成分分析在本质上不可能实现的问题,提出在统计不相关矢量集上进行独立成分分析。与一般白化变换后的数据相比,基于统计不相关矢量集的数据具有更好的不相关性,在估计独立分量时性能更优越。在ORL人脸数据库上进行的实验结果证明了该算法的有效性。     

HTML表格向XML的智能转换

XML已经成为处理与管理信息的标准格式,而HTML表格被广泛应用于Web。为了充分利用与管理HTML表格信息,需要将HTML表格转换成XML。提出一种有效的处理方法,该方法包含2个部分,即表格识别与结构转换。表格识别通过检查格式、语法及语义的特征将表格提取出来并分割成值域与属性域,使用预设的表格模板分析属性域与值域间的层次结构并将其转换成XML格式。通过 300多个表格的实验表明,所提出的方法要优于传统方法,结果的准确率达86.7%。     

文本分类中一种混合型特征降维方法

提出一种基于特征选择和特征抽取的混合型文本特征降维方法,分析基于选择和抽取的特征降维方法各自的特点,借助特征项的类别分布差异信息对特征集进行初步选择。使用一种新的基于PCA的特征抽取方法对剩余特征集进行二次抽取,在最大限度减少信息损失的前提下实现了文本特征的有效降维。对文本的分类实验结果表明,该特征降维方法具有良好的分类效果。