基于聚类核函数的最小二乘支持向量机高光谱图像半监督分类

针对大规模的高光谱数据分类,为了利用未标签样本所含信息,来提升分类器性能,提出了一种半监督分类算法。该算法根据聚类假设,即属于同一类地物的样本点在聚类中被分为同一类的可能性较大的原则来改进核函数,采用基于光谱角度量的K均值聚类算法对样本集进行聚类,根据多次聚类的结果,构造包袋核函数,然后利用加法和乘法运算将包袋核函数和RBF核函数组合成新的核函数,从而把未标签样本信息融入分类器。而且采用最小二乘支持向量机,将标准支持向量机的二次规划问题转换为求解线性方程组的问题。高光谱实测数据实验表明了本文方法的优越性。      

基于运动特征分析的森林火灾烟云实时检测方法

本文提出了一种基于地面视频检测系统中烟云序列图像识别的新方法。利用视频图像的灰度信息,通过运动趋势的估计和相似性的测算,有效提高了具有非结构化及随机性特征的烟云图像检测的可靠性。在仿真试验中,使用此方法对美国Alaska州2005年5月1日拍摄的森林火灾序列图像进行了分析,结果表明它对早期产生的烟云区域能够实时检测和识别,并具有较好的鲁棒性。     

无线传感器网络中一种基于流量预测和相关系数的异常检测方法

几种典型的流量预测模型在无线传感器网络入侵检测系统中的应用进行了分析,结合相关系数,提出了一种异常入侵检测方法.根据WSN节点的预测流量序列和实际流量序列的相关系数变化来进行异常检测.实验结果表明了该方法的有效性,在攻击强度较弱时也具有较高的检测率.     

一种基于森林火险控制的自适应性WSN分簇算法

通过对森林防火现状的总结,分析了WSN在火险控制中的特点,文中提出了一种自适应分簇算法,将整个传感器网络划分成多层.在网络部署初期,为实现监测数据传输的实时性要求,各层簇首节点采用单跳方式与SINK节点直接通信;当节点对环境数据的监测达到一定程度形成预测趋势后,所有的节点可以自适应进入下一个阶段,进行多跳路由的建立,以提高网络的生命周期,延长传感器节点对火场环境监测的时间.     

基于网络摄像机的森林火灾检测系统研究

针对当前世界上发达国家采用卫星作为森林火灾的监测工具,辅以直升机补充卫星监测带的误差,不仅造价高昂,且不符合林区分布广、分布不集中的国情等缺点,提出了一种以网络摄像机作为监测工具,手机、电脑作为处理终端,开发基于Android手机客户端和PC客户端的森林火灾监测系统。利用TCP/IP协议栈设计多线程任务应用,实现火灾监测在网络摄像机平台的移植。最终将视频处理结果由以太网传至监控中心,实现对火灾的提前预警。     

一种基于仿射传播聚类的入侵检测方法

针对基于无监督聚类的入侵检测需要预先指定初始聚类中心和数目的问题,提出了一种基于仿射传播聚类的入侵检测方法,采用了仿射传播聚类实现入侵检测,将每个数据点都看作潜在的聚类中心,通过信息迭代更新自动决定最后的聚类中心和数目,能够获得准确的聚类结果。在对KDD CUP99数据集的仿真实验中验证了方法的可行性,实验结果表明,相比传统方法能有效提高检测率。     

基于分形与特征聚类的DDoS攻击检测方法

针对基于分形的检测方法不深入分析数据包的特点,文中将特征聚类与分形检测有效地集成起来,提出FCCD检测方法,通过对MIT实验室DARPA数据的分析,检验了该方法的有效性。FCCD与单一的分形检测方法相比,在检测到攻击后,可以深入分析出攻击的特征,为数据包的过滤提供依据。FCCD采用了线程同步技术和内存库技术,使分形检测和特征聚类可以高效地并发执行,且只占用较少的系统资源。     

引入负相似的高光谱图像半监督分类

高光谱图像数据体现为波段多、地物标签获取困难大、谱信息抗干扰能力弱等特征,容易引起维数灾难、光谱空间变异性等问题,从而影响分类器的分类精度。针对这些问题,本文将负相似信息引入到拉普拉斯支持向量机（Laplacian Support Vector Machine, LapSVM）的流形正则化项中,提出了一种引入负相似的拉普拉斯支持向量机（Dissimilarity in Laplacian Support Vector Machine, Diss-LapSVM）分类算法,抑制光谱空间变异对分类结果的影响。同时,本文提出利用线性近邻传播（Linear Neighborhood Propagation, LNP）算法构造图的拉普拉斯矩阵,更有效地引入无标签样本的信息。实验结果表明,本文算法的分类精度得到了提高,特别是对光谱特征相似的地物。      

Shift Reduce Parser Based Malicious Sensor Detection for Predicting Forest Fire in WSNs

Nowadays wireless sensor networks (WSNs) has been used in enormous applications for data collection in an unfriendly environment. Forest fire makes vast hazard to the consuming plant of the world. To preserve the forest from fire, sensor nodes monitor the environment temperature. If the temperature is increased and it exceeds the threshold, the sensor sends the notification message to the fire monitoring system. The forest fire is measured by Fire Weather Index. Most of the existing fire monitoring systems only highlight in detection, but not the verification of the sensor. Suppose, if the attacker inserts any malicious sensor, the malicious sensor sends false information or create an additional delay in fire monitoring system. As a result, the more chances for the forest to be destroyed by fire. To solve this problem, Shift Reduce Parser based Malicious Sensor Detection (SRP_MSD) in WSN is proposed. The Bivariate Pascal Triangle method hides the original identity of nodes, data route from the malicious observer and sends confidential information to the Base Station. This method is analyzed and evaluated using network simulator-2. The results show that it is possible to detect malicious sensor nodes and send the reliable information to the forest fire monitoring system.     

A Novel Approach to Unsupervised Change Detection Based on a Semisupervised SVM and a Similarity Measure

This paper presents a novel approach to unsupervised change detection in multispectral remote-sensing images. The proposed approach aims at extracting the change information by jointly analyzing the spectral channels of multitemporal images in the original feature space without any training data. This is accomplished by using a selective Bayesian thresholding for deriving a pseudotraining set that is necessary for initializing an adequately defined binary semisupervised support vector machine classifier. Starting from these initial seeds, the performs change detection in the original multitemporal feature space by gradually considering unlabeled patterns in the definition of the decision boundary between changed and unchanged pixels according to a semisupervised learning algorithm. This algorithm models the full complexity of the change-detection problem, which is only partially represented from the seed pixels included in the pseudotraining set. The values of the classifier parameters are then defined according to a novel unsupervised model-selection technique based on a similarity measure between change-detection maps obtained with different settings. Experimental results obtained on different multispectral remote-sensing images confirm the effectiveness of the proposed approach.     

基于光谱加权直推式支持向量机的高光谱图像半监督分类

高光谱图像分类中的有标签的样本获取较为困难,而半监督分类可以利用到大量未标签样本所含信息,来提高分类准确率。其中直推式支持向量机是标准支持向量机在半监督学习问题上的一种扩展。本文中我们采用凹凸过程规划将直推式支持向量机的非凸目标函数分解为凸函数和凹函数的组合,从而将非凸问题转化为凸优化问题求解。并且针对高光谱图像不同波段鉴别地物类别的能力的差异,为了充分利用各个波段的分类能力,我们引入了光谱权值对支持向量机的核函数进行了改进,对不同的波段赋予不同的权值。实验表明,本文提出的方法在分类正确率以及使用的样本规模上,都表现出了一定的优越性,从而适用于较大规模的高光谱图像分类。      

基于视频图像的森林火灾识别算法研究

针对火焰的色彩特性和燃烧时的动态特性,提出一种面向视频图像序列的火灾识别算法.首先通过火焰颜色特征,在HSV模型基础上运用改进的区域增长法提取目标区域,然后根据火焰区域的重心和面积变化特性,通过设定模糊数学的隶属度函数来判定是否有火灾发生.实验结果表明,本算法实用有效,对火灾的识别率较高.     

基于改进SSA优化WSN分簇协议

无线传感器网络(Wireless Sensor Networks,WSN)的路由协议是无线传感器网络领域中的一个研究热点.针对LEACH协议的不足,提出一种基于自适应t分布改进麻雀搜索算法(Improved Sparrow Search Algorithm,ISSA)的改进LEACH协议(LEACH?ISSA),以解决...     

SC-EEDC: Similarity Based Clustering for Energy Efficient Data Collection in WSN

Wireless Personal Communications - Wireless sensor networks (WSN) are widely used in various situations. Energy saving is one of the most important issues because of the limited power....     

Optimization Based Multi-Objective Weighted Clustering For Remote Monitoring System in WSN

Wireless Sensor Network (WSN) is generally considered as a standout amongst the most critical advancements for the twenty-first century, it normally comprises multifunctional wireless sensor nodes, with detecting, communications, and calculation capacities. Clustering the random nodes in WSN is a challenging task with high performance. This paper presents the new clustering model to monitor the eco-friendly mobile network by clustering the sensor nodes and to enhance the Quality of Service of that optimal network in WSN. The proposed Multi-Objective Weighted Clustering model groups the arbitrary nodes and afterward the optimal network is achieved by the optimization of network parameters. For optimizing the network parameters, a metaheuristic algorithm i.e. Improved Fruit Fly Optimization is introduced. With the goal of assessing the Coverage Efficiency (CE) and network user satisfaction of the accomplished optimal mobile network in WSN, the remote sensor monitoring process is applied. Sensor monitoring helps to know the network users and furthermore to improve the CE of WSN, contrasted with existing work.

     

基于聚类的少数类样本采样方法

针对处理不平衡数据集的分类问题,SMOTE通过在相邻样例间线性插值实现少数类样例过采样。但SMOTE插值的结果是样例密集的地方依然相对密集,样例稀疏的地方依然相对稀疏,影响分类性能。针对该问题本文提出一种基于聚类的过采样方法-C-SMOTE。该方法首先将少数类样例聚成多个簇,再以簇为单位结合SMOTE方法产生新样例。实验结果表明,C-SMOTE既保证了数据集整体分类准确率,又能提高少数类分类精度。     

Improved K-Means Based Q Learning Algorithm for Optimal Clustering and Node Balancing in WSN

Wireless Personal Communications - A wireless sensor network is a potential technique which is most suitable for continuous monitoring applications where the human intervention is not possible. It...     

A Low-Cost Microwave Radiometer for the Detection of Fire in Forest Environments

This paper deals with the development of a microwave noise-adding radiometer, which is purposely designed for the fire detection in forest environments. The sensor operates at 12.65 GHz and exploits commercial Satellite Television (SAT-TV) components such as a parabolic dish and a low-noise block. First, a simple system model is presented to estimate the radiometric contrast due to the presence of fire (increase in the antenna noise temperature with respect to the background) at a certain distance from the receiving antenna. Then, the design of the sensor is addressed, underlining the key technologies that allow the required performance to be attained at low industrial costs. An experimental characterization of the developed radiometer is reported focusing on both accuracy and sensitivity issues. Due to a continuous gain calibration based on the noise-adding procedure, the antenna noise temperature can be retrieved with an absolute error of 4 K without any thermal stabilization of the instrument electronics. Last, real fire detection experiments have been carried out both in laboratory and open-space environments. A radiometric contrast of 8.8 K has been recorded for a wooden fire of 0.2 $hbox{m}^{2}$ placed at a distance of about 30 m from the antenna.