Loading...

Extracting Proper Features for Human Detection in Still Images

Mozafari, Azadeh Sadat | 2011

712 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 42236 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Jamzad, Mansour
  7. Abstract:
  8. Human detection in still images is one of the hardest problems in object detection area. There are several challenges like articulation, pose variation, variant clothing, none uniform illumination, cluttered background and occlusion which make this problem more sophisticated than any other object detection problem. The general solution for this kind of problems is based on supervised learning that contains two main parts: 1-extracting proper features, 2- using proper classifier. The main focus of this thesis is on the first part, extracting proper features, which should be robust to mentioned challenges. Based on the level of extraction we can divide the features to four groups: 1-low-level, 2- mid-level, 3-high-level and 4-hybrid features. In this thesis, we introduced a new set of features which is called Histogram of Small Edges (HOSE) which belongs to midlevel group of features and its base is on template matching and histogram construction that make this set of features robust to pose variation, articulation and cluttered background. However, HOSE doesn’t have better detection rate than state of the art features but it has some properties that can be combined with HOG (a low-level feature) and make hybrid features which can pass the state of the art detection rate. The hybrid features detection rate is 89% TP in 0.02 FP which is 5% better detection rate than other features in 0.02 FP detection rate. In this thesis, all the results are shown on standard INRIA datasets and as the classifier the linear SVM are chosen
  9. Keywords:
  10. Feature Extraction ; Human Detection ; High-Level Features ; Mid-level Features ; Low-Level Features ; Hybrid Features

 Digital Object List

 Bookmark

No TOC