煤层气排采设备视频监控电子稳像系统研究

发布时间：2018-04-22 15:19

本文选题：电子稳像 + 特征点匹配　；参考：《西安科技大学》2017年硕士论文

【摘要】：由于煤层气排采设备的工作环境比较恶劣,监控设备在这种环境中工作容易发生随机的抖动,经常会造成误报警现象。而电子稳像技术可以很好的解决这个问题。该技术是一种稳定视频图像序列的技术,它具有稳像精度高、体积小、重量轻、功耗低等优势。该技术在军事和民用技术都受到了广泛的关注。首先,本文系统的研究了电子稳像基本原理及其处理过程,并对电子稳像算法中四种常用的经典算法进行了分析,其中有块匹配法、位平面匹配法、灰度投影算法、特征向量跟踪法。通过稳像精度、算法的运行速度、算法的适用环境等指标给出了算法各自的优缺点。根据本课题的要求,最终选定用基于Harris角点检测的特征点匹配算法实现电子稳像的目的。在图像预处理的过程中,针对原始的开关中值滤波在保留细线和细节方面的不足,应用了一种改进的开关中值滤波方法。通过仿真实验,对噪声信号从5%增加到30%的图像进行滤波,改进后的方法比原始方法峰值信噪比提高的平均值是3.64ldB,图像的毁坏程度缩小了 1.2%,说明改进的方法可以更好的保留图像中的细线和细节。通过Harris算法对特征点进行提取,并对其进行匹配,从匹配的特征点运动矢量中求得运动参数,根据运动参数对图像进行补偿并输出稳定的图像序列。其次,将电子稳像算法移植到TI公司所生产的硬件平台DM6467上,并对算法进行优化。通过对TMS320DM6467的结构和框架的分析,完成DAVINCI开发环境的配置,并且对算法移植的过程做了详细的说明。通过程序优化和内存的优化两种优化方法对算法进行了优化,提高了算法的运行速度。其中,程序优化中又用了 C语言优化和汇编语言优化。最后,通过主观评价和客观评价两种方式对稳像前后的图像进行质量评价,主观评价给出了图像质量评价的打分方法,它是以多名实验室的同学作为评判员对同一张图像的稳像前后的图像进行打分,然后取平均值,稳像后图像要比稳像前的平均高1.3分左右。应用的客观评价方法包括峰值信噪比,加权峰值信噪比和差影比较法。实验结果显示稳像后的峰值信噪比的值相对于稳像前平均提高了 7.542dB,加权峰值信噪比平均提高了 6.614dB,对比稳像前后相邻两帧图像的差影图可以明显看出稳像后所留下的灰度值小,说明稳定的程度更高。综合两种评价方法得出该算法在硬件平台DM6467上很好的实现了稳像的目的,对稳像的效果比较满意。
[Abstract]:Because of the poor working environment of coal bed methane extraction equipment, random jitter is easy to occur in the monitoring equipment working in this environment, which often results in false alarm phenomenon. And electronic image stabilization technology can solve this problem very well. This technology is a stable video image sequence technology, it has the advantages of high image stabilization accuracy, small volume, light weight, low power consumption and so on. This technology has received extensive attention in both military and civil technology. Firstly, this paper systematically studies the basic principle of electronic image stabilization and its processing process, and analyzes four common classical algorithms of electronic image stabilization, including block matching, bit plane matching and gray projection algorithm. Eigenvector tracking method. The advantages and disadvantages of the algorithm are given by the image stabilization accuracy, the speed of the algorithm and the applicable environment of the algorithm. According to the requirement of this paper, the aim of electronic image stabilization is to use the feature point matching algorithm based on Harris corner detection. In the process of image preprocessing, an improved switching median filtering method is applied to solve the deficiency of the original switching median filter in preserving fine lines and details. Through the simulation experiment, the image whose noise signal is increased from 5% to 30% is filtered. Compared with the original method, the average value of the improved method is 3.64ldB, and the damage degree of the image is reduced by 1.2, which shows that the improved method can better retain the fine lines and details in the image. The feature points are extracted and matched by Harris algorithm. The motion parameters are obtained from the motion vectors of the matching feature points. The image is compensated according to the motion parameters and a stable image sequence is output. Secondly, the electronic image stabilization algorithm is transplanted to the hardware platform DM6467 produced by TI, and the algorithm is optimized. By analyzing the structure and framework of TMS320DM6467, the configuration of DAVINCI development environment is completed, and the process of algorithm porting is explained in detail. Two optimization methods, program optimization and memory optimization, are used to optimize the algorithm and improve the speed of the algorithm. Among them, C language optimization and assembly language optimization are used in program optimization. Finally, the image quality before and after stabilization is evaluated by subjective evaluation and objective evaluation, and the evaluation method of image quality is given. It uses several laboratory students as judges to rate the image before and after the same image stabilization, and then takes the average value. The image after image stabilization is about 1.3 points higher than that before the image stabilization. The objective evaluation methods used include peak signal-to-noise ratio, weighted peak signal-to-noise ratio and contrast comparison. The experimental results show that the peak signal-to-noise ratio (PSNR) after image stabilization is 7.542 dB higher than that before stabilization, and the weighted peak signal-to-noise ratio (PSNR) is increased by 6.614 dB on average. The gray value left behind is small, The degree of stability is higher. By synthesizing two evaluation methods, it is concluded that the algorithm achieves the purpose of image stabilization on the hardware platform DM6467, and is satisfied with the effect of image stabilization.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN948.6;TP391.41

【参考文献】