基于DirectShow的数字电视异态检测系统的设计与实现

发布时间：2018-01-07 09:14

本文关键词：基于DirectShow的数字电视异态检测系统的设计与实现　出处：《电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：在中国广播电视行业的数字化进程不断加快的同时,传统的人工数字电视技术审片模式表现出来的审片耗时长、检测准确度不高、设备花销大等缺点,已经不适合数字电视的审片要求,加之国外审片系统价格高昂,本课题的目的在于研究一个自动技审系统软件,使得其在PC机终端可以表现出来检测速度快、检测精度高、持续审片时间长、成本低等优点,能够改善以上不足。本文就此提出应用计算机软件审片代替人脑大部分审片的自动化数字电视技术审片检测系统,现总结本系统实现采用的方法如下所示:(1)采用DirectShow多媒体技术实现待审节目的音视频流分离,大大方便了软件功能的模块化设计,本次系统集成了源过滤器、MPEG-2解复用器、音频流过滤器、视频流过滤器和音视频播放过滤器。(2)针对常见的视频故障(黑场、彩条、单色场、静帧、YUV超标)进行监测。结合直方图分析数据和故障帧图像特点设计黑场、彩条、单色场、静帧、YUV高标五种故障检测方法,提出将图像等分为大小均一的正方形网格,只针对提取的网格灰度特征进行故障判断,相比较逐个像素点比较判断,网格判断加快软件运行效率。(3)分析MPEG-2视频图像的马赛克故障特点,提出结合差分边缘检测和Canny边缘检测两种方法提取马赛克边缘,并且对提取出来的宏块边缘特征进行分类,还结合宏块灰度特征和宏块色度特征对马赛克故障进行判决。(4)了解音频格式Wave的特点之后,对音频故障(静音、声音过大、声音过小、爆破音、噪声、声音反相)进行检测方法分析和设计。其中,对前四个音频故障分析其在时域上的电平特征来判断故障,采用快速傅里叶变换(FFT)分析噪声和声音反相分别在声音能量和左右声道相位差的特征来判断故障。从系统故障检测的速度、故障检测方法的准确性、系统的稳定性和实时性四个方面出发,对实现的数字电视故障检测系统进行功能性测试,其系统测试结果表明它符合广电行业实际应用的要求,并且其检测准确率基本都高于98%,从而有效地实现数字电视音视频故障检测。
[Abstract]:While the digitization process of China's broadcasting and television industry is speeding up, the traditional technology of artificial digital television film review mode shows the shortcomings of long time consumption, low detection accuracy, large cost of equipment, and so on. It is no longer suitable for the demand of digital TV, and the foreign system is expensive. The purpose of this project is to study a software of automatic technology review system, so that it can be detected quickly in the PC terminal. It has the advantages of high precision, long time of continuous film review, low cost and so on. In this paper, we propose an automatic digital television technology for digital television inspection system, which can replace most of human brain with computer software. This paper summarizes the implementation method of this system as follows: 1) using DirectShow multimedia technology to realize the separation of audio and video stream, which greatly facilitates the modular design of software function. This system integrates source filter MPEG-2 demultiplexer, audio stream filter, video stream filter and audio / video playback filter. Combined with histogram analysis data and fault frame image characteristics, five fault detection methods, black field, color strip, monochromatic field and static frame YUV high standard fault detection method are designed. In this paper, the image is divided into square meshes of uniform size, and fault judgment is only based on the grayscale features of the extracted meshes, and the pixel points are compared to judge each pixel. Grid judgment accelerates the efficiency of software running. (3) analyzing the mosaic fault characteristics of MPEG-2 video image, two methods are proposed to extract mosaic edge, differential edge detection and Canny edge detection. And the extracted macroblock edge features are classified, combined with macroblock gray features and macroblock chromaticity features to determine the mosaic fault. 4) to understand the characteristics of audio format Wave. Analysis and design of detection methods for audio faults (mute, too loud, too low, blasting sound, noise, sound inverse). The level characteristics of the first four audio faults in time domain are analyzed to determine the fault. Fast Fourier transform (FFTFT) is used to analyze the characteristics of noise and sound inversion in sound energy and the phase difference between left and right vocal channels to determine the fault. The speed of system fault detection and the accuracy of fault detection method are analyzed. Based on the stability and real-time of the system, the functional test of the digital TV fault detection system is carried out. The system test results show that the system meets the practical requirements of the radio and television industry. And the detection accuracy is higher than 98, so digital TV audio and video fault detection is realized effectively.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN949.197

【参考文献】