基于稀疏度自适应的矿井智能监控图像重构方法

发布时间：2018-01-30 21:13

  本文关键词： 矿井 智能监控 图像重构 压缩感知 稀疏表示　出处：《煤炭学报》2017年05期 　论文类型：期刊论文

【摘要】：矿井智能监控是实现少人或无人工作面自动化开采和可视化作业的重要保障。针对矿井监控图像易受噪声干扰和雾尘环境等影响,采用传统的基于奈奎斯特(Nyquist)采样和压缩方法存在分辨率低、图像模糊和运算时间过长等问题,根据压缩感知和稀疏重建理论,提出了一种利用分块压缩感知模型和自适应匹配追踪均衡策略获取矿井图像的方法。该方法通过建立矿井图像分块压缩感知模型,信源编码过程先利用稀疏随机矩阵对块图像进行压缩、采样、获得观测值,然后使用DFT作为稀疏基进行信号稀疏表示,最后利用一种改进的自适应匹配追踪算法进行图像重构实现解码。研究结果表明,提出的方法在与其他算法的比较中体现了较好的优越性,能有效提高矿井图像在压缩感知重构阶段的解码质量及其压缩处理速度,具有较强的抗噪声性能和鲁棒性,有助于改善矿井监控图像的清晰度和实时处理性能。
[Abstract]:Mine intelligent monitoring is an important guarantee to realize automatic mining and visual operation of few or no man working face. The image of mine monitoring is vulnerable to noise interference and fog-dust environment and so on. The traditional Nyquist (Nyquist) sampling and compression method based on Nyquist has the problems of low resolution, blurred image and long computation time, according to the theory of compression perception and sparse reconstruction. This paper presents a method of obtaining mine image by using block compression perception model and adaptive matching tracking equalization strategy, which is based on the establishment of block compression perception model of mine image. The source coding process first uses sparse random matrix to compress the block image, samples the observed value, and then uses DFT as the sparse basis for signal sparse representation. Finally, an improved adaptive matching tracking algorithm is used for image reconstruction and decoding. The research results show that the proposed method has better advantages compared with other algorithms. It can effectively improve the decoding quality and compression processing speed of mine image in the stage of compression perception reconstruction. It has strong anti-noise performance and robustness. It is helpful to improve the clarity and real-time processing performance of mine monitoring image.
【作者单位】： 中国矿业大学(北京)机电与信息工程学院;
【基金】：国家重点研发计划重点专项资助项目(2016YFC0801800) 国家自然科学基金资助项目(51674269) 中央高校基本科研业务基金资助项目(2014YJ01)
【分类号】：TD67;TP391.41
【正文快照】： 矿井智能监控、煤岩识别和矿物探测对视频图像的质量和实时性要求较高,目前,在矿井图像压缩方法研究领域,Fourier变换、小波变化及其改进方法被国内外学者广泛研究。文献[1]提出了小波与统计建模的煤岩图像识别方法,取得了较好的识别率与实时性。文献[2]提出了一种双边滤波的，

本文编号：1477242