基于WSN的田野文物入侵侦测系统设计与关键技术研究

发布时间：2018-03-14 20:06

本文选题：文物　切入点：无线传感器网络　出处：《西北工业大学》2014年博士论文　论文类型：学位论文

【摘要】：田野文物具有重要的社会、历史及经济价值。利用先进的技术手段快速有效的获取田野文物环境状态，评价田野文物风险等级，是建立田野文物入侵侦测系统的目的和发展方向。课题根据田野文物保护工作实际需求，提出将无线传感器网络技术应用于田野文物入侵活动侦测中，针对田野文物监测应用的无线传感器网络相关应用面临的应用基础性问题，如侦测对象选择、侦测方法研究、数据压缩及重构、信号特征提取、神经网络性能优化、侦测系统构建、工作流程等问题进行了研究，主要创新点及研究内容如下： 1.从田野文物入侵活动过程及特点出发，结合对现有方法的研究基础上，研究了基于无线传感器网络技术的田野文物入侵活动侦测方法，提出将微地震信号作为侦测对象，并提出了相关的研究技术路线。 2.将压缩感知理论应用于入侵侦测系统，利用压缩感知理论的压缩特性减少系统通信能耗。针对入侵侦测系统数据特点、压缩感知数据重构算法及已有改进算法存在的重构精度差、重构速度慢等不足，提出了基于量子克隆免疫算法的压缩感知数据重构算法Q-CSDR。Q-CSDR对数据进行自适应分帧，并将量子克隆免疫算法应用于压缩感知数据重构过程中以提高数据的重构精度。通过仿真实验结果表明，在稀疏度小于60的条件下，Q-CSDR算法仍能保持85%以上的重构精度，高于其他对比算法。 3.提出了基于低频率采样条件下的信号特征提取算法，，将信号功率谱二次处理算法进行针对性的改进并将其应用于微地震信号特征提取。通过仿真及实际实验结果可以看出，在采样速率为10sps条件下，改进的信号功率谱二次处理算法特征提取性能优于其他特征提取算法，并能够将入侵活动分类精度提高至90%以上。最后分析了算法存在的局限性并提出了高频率条件下的信号采集方法。 4.提出了基于混沌量子克隆免疫的神经网络结构优化方法。通过构建具有一定稀疏度的隐层节点种群来减少神经网络冗余连接，并使用混沌量子克隆免疫算法对各种结构下的神经网络进行寻优，以找到性能最好的神经网络。仿真结果表明，相对于比较算法，本算法具有更好的收敛速度和收敛精度，在逼近函数试验中具有更小的逼近误差。经算法优化后的神经网络能够适应各种地质环境条件下的分类工作，在删除30%以上的隐层冗余节点后分类准确率仍保持在92%以上。 5.设计并实现了基于无线传感器网络的田野文物入侵侦测系统，将入侵活动侦测方法、数据压缩与重构、信号特征提取、分类器性能优化等关键技术应用其中，构建了稳定、可靠的田野文物保护系统。实际实验结果表明，系统能够适应各种地质条件，在分类准确率高于93%的条件下稳定工作2400小时，达到了预期目标。总之，本文研究成果为基于无线传感器网络的田野文物入侵侦测系统的实际应用提供了基础，为推进我国田野文物保护工作的信息化、自动化及智能化提供了一种有效的新思路。
[Abstract]:The field artifacts have important social, historical and economic value. It is the purpose and direction of establishing the field heritage intrusion detection system to acquire the environmental status of the field relics quickly and effectively, and evaluate the risk level of the field cultural relics by using advanced technology.
According to the actual needs of the field of cultural relics protection project, the application of wireless sensor network technology in the field of cultural relics in intrusion detection, aiming at the problem of the basic application of wireless sensor network are related to the application of cultural relics in the field monitoring applications, such as object detection, detection method, data compression and signal reconstruction, feature extraction, neural network performance optimization detection, system construction, work flow and so on, the main innovation and research contents are as follows:
1. from the field of cultural relics invasion process and characteristics, combined with the basic research on the existing methods, the intrusion detection method of wireless sensor network technology based on the cultural relics in the field, the micro seismic signal as the detection object, and puts forward relevant research technical route.
2. the CS theory in intrusion detection system, the compression properties of the compressed sensing theory system for reducing the energy consumption of communication. According to the data characteristics of intrusion detection system, compressed sensing data reconstruction algorithm and improved algorithm of reconstruction accuracy is poor, slow speed of reconstruction is not adequate, proposed quantum clonal immune algorithm for compressed sensing data reconstruction algorithm Q-CSDR.Q-CSDR of adaptive frame data based on quantum cloning and application of immune algorithm in data compression sensing reconstruction process to improve the reconstruction accuracy of data. The simulation results show that the sparsity conditions of less than 60, the Q-CSDR algorithm can still keep the reconstruction accuracy of more than 85%, higher than the other compared algorithms.
3. the low frequency signal feature extraction algorithm based on sampling condition, the power spectrum of the signal processing algorithm for two times improvement and its application in micro seismic signal feature extraction. Through simulation and actual experimental results show that the sampling rate of 10sps under the condition of extraction performance than other feature extraction algorithm of signal the power spectrum improved two processing algorithm characteristics, and can improve the classification accuracy of intrusion activities to more than 90%. The final analysis of the limitations of the algorithm are proposed and the signal acquisition method of high frequency conditions.
4. proposed neural network structure optimization based on Chaos Quantum immune clone. Through constructing a hidden node population must sparsity to reduce redundant network connectivity, and neural network structure to the optimization using the Chaos Quantum clonal immune algorithm and neural network to find the best performance. The simulation results show that compared with the the comparison algorithm, the convergence speed and precision of this algorithm has better approximation function, in the test with the approximate error is smaller. The neural network optimized algorithm can adapt to the classification of various geological conditions, the accuracy rate remained at more than 92% in the hidden layer to delete redundant nodes after the classification of more than 30%.
5. the design and implementation of wireless sensor network intrusion detection system based on the field of cultural relics, the intrusion detection method, data compression and signal reconstruction, feature extraction, classifier performance optimization and other key technology application, to construct stable, reliable protection of cultural relics in the field system. The experimental results show that the system can adapt to various geological conditions, accurate the classification rate is higher than 93% under the condition of stable work 2400 hours, to achieve the expected goal.
In conclusion, the research results provide a basis for the practical application of the field heritage intrusion detection system based on wireless sensor networks, and provide an effective new way to promote the informatization, automation and intellectualization of the field cultural relic protection in China.

【学位授予单位】：西北工业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP393.08

【参考文献】