Dissertation > Excellent graduate degree dissertation topics show

Study on Detection and Recognition of Logos in Video Stream

Author: MaoYunFeng
Tutor: ZhangXianMin
School: Shanghai Jiaotong University
Course: Pattern Recognition and Intelligent Systems
Keywords: edge operator correlation similarity logo detection logo recognition
CLC: TP391.41
Type: Master's thesis
Year: 2009
Downloads: 17
Quote: 0
Read: Download Dissertation

Abstract


A variety of video productions contain logos, which as a specific semantic symbol indicate the information about video contents, so they have a great potential for study and application in video processing fields.This paper mainly discusses the fundamental theories and key technologies of detection and recognition of still logo in video, and proposes an application framework. The major theoretical part consists of object detection, edge operators, filtering methodologies, image matching, common image features and performance evaluation methods. This paper employs these technologies in the real time surveillance field.After analyzing conventional detection algorithms, combining characteristics of TV video stream, we use short interval inter-frame variance method to get logo mask, divide logos into two classes——opaque and semi-transparent——to process respectively, and then choose proper frame to mask. The quality of extracted logo will affect processes afterward, so in this phase we use much priori knowledge of TV logo.In feature selection phase, the key problem is which representative feature to select in order to better separate different TV logos. TV logos always have a simple structure, clear colors and distinct shapes. And for semi-transparent logos, edge is much easier and more stable to extract, so we choose logo edge as its feature. This paper use Laplacian edge operator with filtering method to spawn logo template, compared with other common edge operators like Sobel, Canny.In real-time logo recognition phase, we extract Laplacian edge image of logo region in video stream regularly, match it with corresponding TV logo template in the logo template database by correlation similarity metric, and finally make a decision to achieve identification or recognition. Optimal empirical thresholds derive from real experiments, and we also test multiple times to judge for semi-transparent logo. At last according to experimental data under system requirements, we get false acceptance rate and false rejection rate and, analyze the results and make some improvement.The scheme of detection and recognition of still logo in video proposed in this paper has been applied to automatic monitoring system of TV signals, and the resulting data demonstrate this approach gains desirable recognition rate and is applicable to the real application.

Related Dissertations

  1. Profile Measuring System Based on Digital Image Processing,TP274
  2. The Correlation Analysis Based on the Fuzzy Similarity of Property,TP393.08
  3. Research on Dairy Cow Insurance by Digital Image Analysis and Recognition Technology,TP391.41
  4. Research on Algorithm of Moving Vehicle Logo Recognition,TP391.41
  5. Research on PCA and Sample Fusion Based Vehicle Logo Recognition,TP391.41
  6. Logo Recognition,TP391.41
  7. Homology Theory and Complex of Polynomial Ideal with Several Generators,O153.3
  8. Study on Methods for License Plate Detection and Car Type Classification,TP391.41
  9. Research on Color Calibration and Recognition Algorithm of Robots in Vision System of MiroSot,TP242
  10. The scene edge extraction technology research and application,TP391.4
  11. Positioning problem with radial identify skeletal age identification system,R318
  12. Research for Recognition of Moving Vehicle Type,TP391.4
  13. Research of Type-Recognition Techniques for Moving-Vehicles,TP391.4
  14. Research of Recognition Techniques for Moving Vehicles,TP391.4
  15. Fault Management Based on Event Correlation and Data Mining,TP393.07
  16. Vehicle Type Automatic Recognition System,TP391.41
  17. Study of Several Technologies on Vehicle License Plate and Logo Recognition,TP391.41
  18. The Studies on Structural Damage Identification Method Based on Edge Operator,TP391.41
  19. Pressure Equalization of the Rain-Screen Wall under Wind Excitation,TU227
  20. Study on Image Mining Based on Formal Concept Analysis,TP391.41
  21. The Research for Recognition of Moving Vehicle,TP391.4

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile