Research On Caption Detection Technology Supporting Video Intelligence Analysis

In the era of information, intelligence which was a indemnification factor of traditional war and strategic rivalry is becoming a primary factor of warring, and it has a great influence on decision-making. With the developing of multimedia technology, digital storage technology and computer technology, video is becoming one of the most important source of intelligence. Thus it becomes more and more urgent to extract valuable intelligence from large video data sets, and one of the major challenges is finding effective analytic tools and methods. This thesis probes on the caption detection technology which assists video analysis. The main work and contribution of this paper is as follows:We have studied the application of caption detection to assisting video intelligence processing and analysis; caption detection assisting structuredness of video; caption detection assisting fabrication of video abstract; caption detection and other multimedia characteristic assisting classification of video and using caption to assist fabrication of index of video semantic.It has studied the caption frame detection technology. After analyzing the special texture of caption frame and the appearance or disappearance characteristic of caption event, we have put forward two algorithms of detection of caption frame: Compressed-domain Texture Energy-Based method (CEB method), Edge Variety-Based method (EVB method) and evaluated the performance of the algorithms by some experiments.Study to caption detection technology in single frame video image. After analyzing the shortage of proposed algorithm, we have put forward two algorithms to detect caption area in single frame: Block Intensity Detection-Based caption detection algorithm (BTB method); Multi-Frames Integration-Based caption detection algorithm (MFIB method). We have evaluated the performance of algorithms by experiments in the end.We have designed and realized caption detection system. This system not only realized the caption detection algorithms proposed in this paper, but also validated the application of caption detection to assisting video intelligence processing and analysis-by fabricating video abstract and assisting video annotation.