基于DFT基的矿井视频监控图像分块压缩感知方法

针对矿井视频监控图像受噪声干扰影响大,采用常规的图像采样和压缩方法存在图像模糊和传输时间过长等问题,提出了一种矿井视频监控图像分块压缩感知方法。该方法通过建立矿井视频监控图像分块压缩感知模型,在井下图像采集节点利用稀疏随机矩阵进行压缩采样,然后在地面监控中心利用正交匹配追踪( OMP )算法重构图像。研究结果表明,采用本文算法的重构图像误差小、重构时间短,所需信号采样点数少；与扰频Hadamard矩阵相比,采用稀疏随机矩阵和高斯随机矩阵作为观测矩阵对图像信号重构的峰值信噪比( PSNR)提高4 dB~5 dB；本文算法与基于小波基的算法相比,信号重构的PSNR提高1 dB~4 dB,重构时间缩短至少80％以上。

The block compressed sensing of mine monitoring images based on DFT basis

To address the problem that the captured videos exist low resolution with noise and long transmission time by using conventional methods of images sampling for mine videos,based on compressed sensing,the algorithm of block compressed sensing for mine videos is proposed. By establishing model of block compressed sensing in mine monitoring videos,the proposed method uses sparse random matrix to sample mine images on sensing nodes. Then,it employs orthogonal matching pursuit ( OMP ) algorithm to reconstruct image on monitoring center. The results indicate that the proposed method compares favorably with existing schemes at lower reconstruction error, shorter reconstruction time and less sampled data.The Peak Signal￣to￣Noise Ratio(PSNR)of the algorithm is 8 dB~10 dB higher than that of the method using Scrambled Hadamard matrix,and simultaneously is improved by 1 dB~4 dB in comparison with that of the algorithm which base on wavelet basis,but the time is shortened at least 80%.