顧孟愷臺灣大學：資訊工程學研究所許傅鈞Hsu, Fu-ChunFu-ChunHsu2007-11-262018-07-052007-11-262018-07-052007http://ntur.lib.ntu.edu.tw//handle/246246/53619Structural analysis of video is an essential step to automatic video content analysis. In a general video structure, shot is the most basic unit, and it must be determined before any multimedia content-based retrieval applications. However, despite the rich research efforts in software shot detection system, there is a lack of shot detection system that is designed in hardware. In this thesis, we proposed a shot boundary detection system using hardwarebased Support-Vector-Machine(SVM). Our system can detect both cut shots and gradual shots. We optimized the feature extraction process based on global color histogram with a pipelined architecture to save memory and increase the overall speed. The class imbalance problem in shot detection is solved by using random pseudo-sampling at the SVM training stage. The digital hardware SVM is designed using a fully-parallel pipelined architecture and is highly configurable on vector dimensions. Our data wordlength only used 4 bits signed integer plus one bit decimal in fixed number, while the detection accuracy is competitive comparing with floating point software. Our SVM classifier presented here can run a speed up to 251.62MHZ on Xilinx Virtex IV XC4VSX35 FPGA, and the video processing on our shot detection system can achieve 128 fps on PC. That makes our system met the real-time constraints, and detect shots in a continous video stream such as live-TV news or sports game.1 Introduction 1 1.1 Overview or shot boundary detection System . . . . . . . . . . 1 1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3 Our Contribution . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.4 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . 9 2 Shot Boundary Detecton System and SVM 10 2.1 Introduction of the Shot Boundary Detection System . . . . . 11 2.1.1 Features Extraction . . . . . . . . . . . . . . . . . . . . 12 2.1.1.1 General Features ExtractionMethods . . . . 13 2.1.1.2 Temporal Kernel Correlation Filtering . . . . 15 2.2 Support Vector Machine . . . . . . . . . . . . . . . . . . . . . 17 2.2.1 Brief Introduction . . . . . . . . . . . . . . . . . . . . . 17 2.2.2 Shot boundary classification using SVM . . . . . . . . 19 2.2.3 Hardware Implementation of SVM . . . . . . . . . . . 20 2.2.4 Hardware Shot Detector . . . . . . . . . . . . . . . . . 21 3 System and Hardware Architecture 22 3.1 Feature Extraction in Shot Boundart Detection . . . . . . . . 23 3.1.1 Visual Representation of Video . . . . . . . . . . . . . 23 3.1.2 Continuity FeatureMeasurement . . . . . . . . . . . . 24 3.1.3 A Pipelined Temporal Kernel Correlation Architecture 25 3.2 Support Vector Machine Architecture . . . . . . . . . . . . . . 26 3.2.1 Proposed Hardware Architecture . . . . . . . . . . . . 28 3.2.1.1 Hardware design platform . . . . . . . . . . . 28 3.2.1.2 Class Imbalance in Shot Boundary Detection 29 3.2.1.3 A Fully-Parallel Pipelined SVM . . . . . . . . 30 3.2.2 Post-processing Refinement of Shot Detection System . 33 4 Experimental Result 35 4.1 EvaluationMetrics . . . . . . . . . . . . . . . . . . . . . . . . 35 4.2 Training using RandomPseudo-Negative sampling . . . . . . . 36 4.3 Hardware Synthesizing Results . . . . . . . . . . . . . . . . . . 38 4.3.1 Result compares with other SVMhardwares . . . . . . 38 4.3.2 Result compares with other hardware shot detector . . 39 4.4 Wordlength Quantization . . . . . . . . . . . . . . . . . . . . . 40 4.4.1 Discussion on Effects of Quantization in Shot Detection 42 4.5 Performance of the Proposed Shot Detection System . . . . . 44 4.5.1 Realtime Shot Detection on Various Video Types . . . 45 5 Conclusion 48 5.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 5.2 FutureWork . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49805799 bytesapplication/pdfen-US支援向量機硬體影片分段偵測Support Vector Machinehardwareshot boundary detectionthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/53619/1/ntu-96-R94922082-1.pdf