主成分分析在均值漂移遥感影像分割中的应用

针对均值漂移分割算法假设样本处于欧式空间,而遥感影像由于波段范围广,不能像普通彩色图像一样,能够由RGB空间转换到LUV空间来满足这种假设的问题,该文提出采用主成分分析(Principal Component Analysis,PCA)对遥感影像进行变换,以满足这种假设。对于输入的一幅遥感影像,首先采用PCA对该影像进行变换,并且将变换后的数据线性拉伸到0~255之间,然后替换原始影像数据,并调用均值漂移算法进行影像分割。实验结果表明,采用所提出方法得到的结果优于直接对原始影像进行均值漂移分割的结果。     

基于显著性密度修正的均值漂移分割算法

针对固定空间和色彩带宽的均值漂移分割算法无法解决的错分割问题,提出一种基于显著性特征进行密度修正的均值漂移分割算法。首先基于密度估计的主颜色量化结果计算区域视觉显著性;其次,将区域视觉显著性融合像素级显著性作为色彩特征空间聚类的密度修正因子,将密度修正后的融合图像作为输入执行均值漂移分割;最后进行小区域合并获得最终分割结果。实验结果显示,所提分割算法在四种尺度上的真实边界准确率和召回率平均值达到0.64和0.78,与其他方法相比,分割精度有显著的提高;同时,在视觉上有效提高了目标完整性,增强了自然图像中目标分割的鲁棒性。     

基于双重超像素集的快速路径相似度图像分割算法

为克服基于路径相似度计算时间复杂度高以及基于单一过分割区域集的聚类方法 容易导致误合并的缺陷, 提出一种结合均值漂移和路径相似度的谱聚类算法. 该算法使用超像 素构建基于路径相似度的模型来实现加速. 首先, 利用均值漂移算法对图像进行两次预分割(不同参数), 将这些过分割区域视为两组超像素集合, 构建基于双重过分割区域集的加权图; 之后, 使用各超像素的色彩均值和超像素间存在的交叉像素计算初始相似度, 再利用路径相似度模型得 到基于路径的相似度; 最后, 采用Multiway Ncut算法进行聚类. 通过算法自身参数和图结构实验, 测试算法的鲁棒性和稳定性; 通过多幅彩 色图片的分割实验, 表明本文的方法在准确性和时效性方面都具有很好的性能.     

区域化模糊均值高分辨率彩色遥感影像分割方法

基于像素模糊?? 均值算法(FCM) 及其改进算法难以解决高分辨率遥感影像中地物目标光谱测度相似性减弱和几何噪声增大带来的分割难题, 提出一种基于区域的FCM算法. 该方法利用Voronoi 几何划分将影像域划分为子区域, 并用子区域拟合地物目标的几何形状. 在此基础上, 定义区域FCM目标函数, 通过迭代最小化该目标函数实现高分辨率遥感影像分割. 实验结果表明, 与基于像素的FCM和增强FCM方法相比, 所提出方法可以更加精确地实现高分辨率遥感影像分割.

     

融合颜色和区域信息的彩色图像分割新方法

结合了均值漂移算法和区域合并算法,取长补短,提出了一种融合颜色和区域信息的彩色图像分割方法。该算法首先利用均值漂移求取各个局部极值(聚类中心),在带宽求取和权重设置上使用了自适应法则,使算法更具有适用性;然后使用一个基于阀值的区域合并算法,解决了均值漂移对纹理和关照变化的过分割。实验证明,该算法是有效的。     

基于空间相关性的图像分割算法研究

提出一种充分利用图像的空间相关性来达到高效快速地进行图像分割的新方法。利用均值漂移算法对图像进行分割形成过度分割的区域,并使这些区域保持理想的边缘和空间相关部分,用图结构表示的区域相邻图来代替分割的区域。和K-均值算法的思想一样,迭代循环置信传播算法以其具有收敛速度快的特点被用于最小化开销函数、整合过度分割的区域和获得最终的分割结果。基于分割区域而不是图像像素的图像聚类分割方法可降低噪声敏感性,同时提高图像分割质量。与FCM和MRF算法相比较,该算法在复杂场景图像中显示了更好的分割性能。     

尺度自适应的高分辨率遥感影像分水岭分割方法

高空间分辨率遥感影像的分割是面向对象分析的重要基础,大部分影像分割算法都涉及分割参数设置的问题,影像分割方法的参数自适应是影响遥感影像分割效率和效果的关键问题之一。针对传统分水岭分割算法易受噪声干扰且分割尺度参数难以自适应选择的问题,提出了一种尺度自适应的高分辨遥感影像分水岭分割方法,即在对原始影像进行中值滤波的基础上,综合考虑高分遥感影像多个波段信息,并利用区域合并进行分水岭分割,然后在空间统计学思想下,统计不同窗口下局部方差的变程实现分水岭分割参数的自适应设置,进而对高空间分辨率遥感影像进行分割。最后,以IKONOS和QuickBird高空间分辨率多光谱遥感影像作为实验数据,对提出参数自适应的分水岭分割方法的有效性进行了验证。基于分割后斑块内均质性和斑块间异质性指标构建综合评价模型,对本文提出的分割方法所得到的分割结果与不同参数序列的分割结果进行了定量比较,对比结果表明采用该分割算法能够得到较好的分割效果。因此,该方法不仅能一定程度上提高影像分割的精度,也保证了分割参数选择的自动化程度,可为今后的影像分割及参数化研究提供一种思路。     

基于均值漂移和区域合并的指针仪表图像分割

均值漂移算法在图像分割中有着十分重要的作用,考虑到图像的噪声和边缘保持等因素,提出了一种基于均值漂移的图像平滑和基于阈值的区域合并的指针式仪表图像分割方法.首先利用均值漂移算法对图像进行滤波降噪,然后对滤波降噪后的图像通过阈值的区域合并方法进行阈值分割.经过均值漂移滤波后的图像不仅能很好地保持指针仪表图像的边缘特征,还能有效地去除噪声,而基于阈值的区域合并方法能够使得分割的图像效果更明显.     

一种有效的遥感图像无缝分割方法

遥感图像具有数据量大、模糊性较强、纹理细节丰富等特点，这就决定了遥感图像的分割与自然景色图像分割相比，在分割效率、分割效果上都提出了更高的要求。另外，无缝空间遥感影像数据库的建立为我们提出了在基于内容的影像检索中遥感图像无缝分割问题。在JSEG算法的基础上，通过图像增强技术，解决了遥感图像不同区域边界往往不明显的问题，大大改善了分割效果；为了解决分割的无缝性，提出了基于gabor纹理特征的区域边界合并算法。实践证明本方法很好地解决了遥感图像分割的无缝性问题。     

基于图割与概率图谱的肝脏自动分割研究

提出一种利用均值漂移算法做预处理,将概率图谱与图割算法相结合的肝脏自动分割方法。该方法一方面有效利用概率图谱所代表的形状信息,并且以无参形状建模;另一方面,均值漂移算法对灰度和形状信息组成的综合信息进行过分割处理,利用过分割的区域代替单个像素参与图割算法,降低了优化算法的复杂度。实验结果表明,该方法有效结合了概率图谱和均值漂移算法的优点,提高了图割算法的精度和速度。     

3D human motion tracking based on a progressive particle filter

Human body tracking has received increasing attention in recent years due to its broad applicability. Among these tracking algorithms, the particle filter is considered an effective approach for human motion tracking. However, it suffers from the degeneracy problem and considerable computational burden. This paper presents a novel 3D model-based tracking algorithm called the progressive particle filter to decrease the computational cost in high degrees of freedom by employing hierarchical searching. In the proposed approach, likelihood measure functions involving four different features are presented to enhance the performance of model fitting. Moreover, embedded mean shift trackers are adopted to increase accuracy by moving each particle toward the location with the highest probability of posture through the estimated mean shift vector. Experimental results demonstrate that the progressive particle filter requires lower computational cost and delivers higher accuracy than the standard particle filter.     

Real-time hand tracking using a mean shift embedded particle filter

Particle filtering and mean shift (MS) are two successful approaches to visual tracking. Both have their respective strengths and weaknesses. In this paper, we propose to integrate advantages of the two approaches for improved tracking. By incorporating the MS optimization into particle filtering to move particles to local peaks in the likelihood, the proposed mean shift embedded particle filter (MSEPF) improves the sampling efficiency considerably. Our work is conducted in the context of developing a hand control interface for a robotic wheelchair. We realize real-time hand tracking in dynamic environments of the wheelchair using MSEPF. Extensive experimental results demonstrate that MSEPF outperforms the MS tracker and the conventional particle filter in hand tracking. Our approach produces reliable tracking while effectively handling rapid motion and distraction with roughly 85% fewer particles. We also present a simple method for dynamic gesture recognition. The hand control interface based on the proposed algorithms works well in dynamic environments of the wheelchair.     

基于MAP估计的直接均值移动归一化训练方法

提出了一种基于数学统计模型化的说话人归一化训练方法，它将与状态相关的直接均值移动归一化训练方法和MAP／WNR模型自适应方法结合到统一的鲁棒性框架中，为模型自适应方法提供了更加合适的初始模型，在提高自适应速度和保持足够的模型平滑度之间得到了较好的平衡，实验表明，该方法可有效的提高有监督模式下语音识别的鲁棒性。     

基于结构特征的飞机目标识别方法

针对高分辨率可见光图像的飞机目标智能识别,提出一种基于飞机结构特征的识别方案。采用Mean shift方法分割图像,提取区域边界,得到目标轮廓集,对轮廓采取多种结构化表示方式,便于分析轮廓的特征及相互关系;综合利用尺寸、占空比、长宽比、凸包、直线段等多个形状特征筛选轮廓,定位飞机目标,确定飞机轴向,获取目标几何参数。实验结果证明,该方法定位精确、识别率高,是一种快速、稳健的飞机识别方案,对类似的人工目标的识别具有一定的参考意义。     

RADARSAT for oil spill trajectory model

This work introduces a new approach for utilizing RADARSAT data for oil slick detection and oil slick trajectory model in the coastal water of Malacca Straits. The concept of the study involves two sub-models. The first model contains texture algorithms for oil slick automatic detection. The second model contains the oil slick trajectory forecasting model. The oil slick trajectory model contains the integration between Doppler frequency shift model and Lagrangian model. Doppler frequency shift model is implemented to simulate tidal current pattern from RADARSAT data, while the Lagrangian model is used to predict oil slick spreading pattern. The classical Fay’s algorithm was implemented in the two models to simulate the oil slick trajectory movement.The results of this study are for a specific RADARSAT image in Malacca Straits and not a generalization for all oil slick incidents. The study shows that the entropy and homogeneity texture algorithms are appropriate for oil slick detection. The results shows that oil slick length is effected by the tidal current V component which had a maximum velocity of 1.4 m/s. The tidal current V component moved towards the north direction. The oil slick trajectory path is moved towards the coastline due to the effect of the tidal current U component. The oil slick approached the coastline within 48 h.In conclusion, RADARSAT HH polarization data can image the oil slick by using the entropy texture algorithm. The forecasting of oil slick trajectory movement can be detected by the integration of Fay’s algorithm model and Doppler frequency shift model.     

复杂运动目标的学习与识别

针对复杂运动目标识别问题,提出了一个基于反馈型随机神经网络的运动认脸与物体的自动识别系统,该系统具有强大学习能力,运动目标检测与识别快速准确等特点,在对该的核心－反馈型二元网络进行深入分析的基础上,提出了一种适合于该神经网络模型的高效渐进式Boltzmann学习算法,实验结果表明,该识别系统性能优异,在几个方面超过了eTrue公司的TrueFace人脸识别系统。     

Identification of control chart patterns using wavelet filtering and robust fuzzy clustering

This paper proposes a hybrid framework composed of filtering module and clustering module to identify six common types of control chart patterns, including natural pattern, cyclic pattern, upward shift, downward shift, upward trend, and downward trend. In particular, a multi-scale wavelet filter is designed for denoising and its performance is compared to single-scale filters, including mean filter and exponentially weighted moving average (EWMA) filter. Moreover, three fuzzy clustering algorithms, based on fuzzy c means (FCM), entropy fuzzy c means (EFCM) and kernel fuzzy c means (KFCM), are adopted to compare their performance of pattern classification. Experimental results demonstrate that the excellent performance of EFCM and KFCM against outliers, especially in the case of high noise level embedded in the input data. Therefore, a hybrid framework combining wavelet filter with robust fuzzy clustering is suggested and proposed in this paper. Compared to neural network based approaches, the proposed method provides a promising way for the on-line recognition of control chart patterns because of its efficient computation and robustness against outliers.     

自适应带宽均值移动算法及目标跟踪

首先提出了一种经典均值移动算法的推广算法,即自适应带宽均值移动算法,进而提出了基于自适应带宽均值移动的二维视频目标跟踪算法（ABMSOT）.前者提出了在带宽自适应情况下均值移动算法求取局部极值的框架步骤,后者可实时跟踪目标的位置、大小和方向.在ABMSOT算法中,目标模型和候选模型采用自适应带宽核函数加权特征直方图描述,目标模型和候选模型的相似性采用Bhattacharyya系数度量;通过迭代两步法搜索到目标最有可能的位置、大小和方向.第一步执行一次均值移动迭代搜索目标位置,第二步计算出最能描述目标区域大小和方向的带宽矩阵.从理论上证明了两个算法的收敛性,并通过实验证明了ABMSOT算法能实时跟踪目标的位置、大小和方向.     

Mean shift, mode seeking, and clustering

Mean shift, a simple interactive procedure that shifts each data point to the average of data points in its neighborhood is generalized and analyzed in the paper. This generalization makes some k-means like clustering algorithms its special cases. It is shown that mean shift is a mode-seeking process on the surface constructed with a “shadow” kernal. For Gaussian kernels, mean shift is a gradient mapping. Convergence is studied for mean shift iterations. Cluster analysis if treated as a deterministic problem of finding a fixed point of mean shift that characterizes the data. Applications in clustering and Hough transform are demonstrated. Mean shift is also considered as an evolutionary strategy that performs multistart global optimization     

结合粒子滤波与Mean shift的高速运动目标跟踪

Mean shift跟踪算法能够有效跟踪视频序列中的各种运动目标,但是该算法无法准确地跟踪视频中高速运动目标.通过分析mean shift算法的原理,指出mean shift对高速运动目标跟踪失效的原因,提出一种基于mean shift的粒子滤波跟踪的新算法.通过实验比较,该算法能改善了Mean shift算法对高速运动目标的效果,并且在存在干扰目标的情况下具备良好的跟踪效果.