面向山区铁路异物侵限监测的压缩感知视频图像处理方法研究

Most of the railways are located in the mountainous areas in our country. Due to its frequent occurrence, railway foreign invasion has become a big threat to the traffic safety. With the ability of perceiving rich multimedia information and realizing a full range of video monitoring, Wireless Multimedia Sensor Networks is highly suitable for railway foreign invasion monitoring, while it is challenged by the limitation of data transmission and network energy..Based on the research of characteristics of geographical environment in the critical areas of mountainous railways, this project firstly employs the theory of compressed sensing and completes the sparse representation about the characteristics of video image to decrease transferred data and prolong the life-cycle of WMSN. Taking the sparsity as the foundation, the combined method of video images fusion, denoising and reconstruction is studied to obtain high-quality video images to monitor railway foreign invasion in mountainous areas. At last, the simulated experiment system will be established to verify and optimize the achievement of the study. The project aims to combine the compressed sensing with video image processing so as to get high quality video images..When this project is accomplished, it can provide a theoretical basis and technical support for railway foreign invasion monitoring and video image processing in WMSN. Furthermore, this project will lay a solid foundation to realize real-time monitoring of railway foreign invasion all-weather and all-around.

我国大部分铁路都分布于山区地带，而山区铁路异物侵限事件频发，对铁路安全运行造成重大威胁；无线多媒体传感器网络感知信息丰富，能实现全方位视频监测。将无线多媒体传感器网络应用于山区铁路异物侵限监测的优势明显，但存在着传输数据量大且网络能量受限的局限性。.本项目在深入研究山区铁路异物侵限多发地段场景特征的基础上，借鉴压缩感知理论，对场景视频图像特征进行充分的稀疏表示，以减少处理的数据量、延长网络生命周期；在稀疏表示的基础上，研究视频图像融合、去噪及重构方法，以获取高质量的视频图像、实现山区铁路异物侵限可靠监测；建立仿真实验系统，实现对研究成果的验证和优化。本项目旨在将压缩感知和视频图像处理充分结合，有效地获取高质量视频图像。.项目完成后，将为无线多媒体传感器网络中视频图像处理及山区铁路异物侵限监测提供理论依据与技术支持，为实现针对山区铁路异物侵限多发地段的全天候、全方位实时监测奠定良好基础。

无线多媒体传感器网络（Wireless Multimedia Sensor Networks，WMSN）具有布线方便、感知信息丰富、监测范围大等优点，适用于山区铁路异物侵限监测，但存在着传输数据量大且网络能量受限的局限性。本项目针对山区铁路异物侵限多发地段的复杂场景，通过分析该区域视频图像特征，构建了视频图像的特征模型。同时，借鉴超完备信号稀疏分解和非线性逼近理论，根据红外、可见光视频图像的不同成像特点及其时空相关性，构造了由混合正交基函数组成的字典，实现了对视频图像的稀疏表示。此外，设计了自适应加权平均融合算法，该算法分别将关键帧及残差图像的观测值进行融合，获取了融合观测值。在此基础上，利用图像的稀疏性和非局部自相似两种先验信息构建了稀疏正则化去噪模型，并针对该模型，提出了将迭代收缩阈值和全变分相结合的模型求解算法，在提高视频图像质量以实现山区铁路异物侵限可靠监测的同时，显著减少传输的数据量，延长了WMSN的网络生命周期。本项目共发表科技论文22篇（其中SCI 3篇、EI 12篇、国际会议3篇、国内核心4篇），申请国家发明专利6项，获得3项授权实用新型专利；培养硕士研究生（已毕业）12名。

内容获取失败，请点击重试