recognition and tracking of human body parts

30
Recognition and tracking of human body parts Algirdas Beinaravičius Gediminas Mazrimas Salman Mosslem

Upload: gilda

Post on 15-Feb-2016

46 views

Category:

Documents


0 download

DESCRIPTION

Recognition and tracking of human body parts. Algirdas Beinaravičius Gediminas Mazrimas Salman Mosslem. Contents. Introduction Background subtraction techniques Image segmentation Color spaces Clustering Blobs Body part recognition Problems and conclusion. Introduction. Project tasks. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Recognition and tracking of human body parts

Recognition and tracking of human body parts

Algirdas BeinaravičiusGediminas Mazrimas

Salman Mosslem

Page 2: Recognition and tracking of human body parts

Introduction Background subtraction techniques Image segmentation

◦ Color spaces◦ Clustering

Blobs Body part recognition Problems and conclusion

Contents

Page 3: Recognition and tracking of human body parts

Background subtraction/Foreground extraction

Color spaces and K-Means clustering Blob-level introduction Body part recognition

Introduction. Project tasks

Page 4: Recognition and tracking of human body parts

What is background subtraction? Background subtraction models:

◦ Gaussian model◦ “Codebook” model

Background subtraction

Page 5: Recognition and tracking of human body parts

Learning the model Gaussian parameters estimation

Thresholds - Foreground/Background determination

Background subtractionGaussian model

Page 6: Recognition and tracking of human body parts

Background subtraction“Codebook” model

Page 7: Recognition and tracking of human body parts

Background subtractionModel comparison

Page 8: Recognition and tracking of human body parts

Color spaces◦ RGB◦ HSI◦ I3 (Ohta)◦ YCC (Luma Chroma)

Clustering◦ K-Means◦ Markov Random Field

Image segmentation

Page 9: Recognition and tracking of human body parts

RGB (Red Green Blue)◦ Classical color space◦ 3 color channels (0-255)

In this project:◦ Used in background subtraction

Image segmentationColor space: RGB

Page 10: Recognition and tracking of human body parts

HSI (Hue Saturation Intensity/Lightness)◦ Similar to HSV (Hue Saturation Value)◦ 3 color channels:

Hue – color itself Saturation – color pureness Intensity – color brightness

◦ Converted from normalized RGB values◦ Intensity significance minimized

In this project:◦ Used in clustering◦ Blob formation◦ Body part recognition

Image segmentationColor space: HSI

Page 11: Recognition and tracking of human body parts

Image data (pixels) classification to distinct partitions (labeling problem)

Color space importance in clustering

Image segmentationClustering

Page 12: Recognition and tracking of human body parts

Clustering without any prior knowledge Working only with foreground image Totally K clusters Classification based on cluster centroid and

pixel value comparison◦ Euclidean distance:

◦ Mahalanobis distance:

Image segmentationClustering: K-Means

Page 13: Recognition and tracking of human body parts

Image segmentationClustering: K-Means Euclidean/Mahalanobis distance comparison

Euclidean distance Mahalanobis distance

Page 14: Recognition and tracking of human body parts

Image segmentationClustering: K-Means color space comparison

RGB HSI

Page 15: Recognition and tracking of human body parts

Probabilistic graphical model using prior knowledge

Usage:◦ Pixel-level◦ Blob level

Concepts from MRF:◦ Neighborhood system◦ Cliques

Image segmentationClustering: MRF

Page 16: Recognition and tracking of human body parts

Image segmentationClustering: MRFNeighborhood system

Cliques

Page 17: Recognition and tracking of human body parts

Blob parameters Blob formation Blob fusion conditions Blob fusion

Blobs

Page 18: Recognition and tracking of human body parts

Higher level of abstraction◦ Ability to identify body parts◦ Faster processing

Blobs

Page 19: Recognition and tracking of human body parts

Label. Set of area pixels. Centroid. Mean color value. Set of pixels, forming convex hull. Set of neighboring blobs. Skin flag.

BlobsParameters

Page 20: Recognition and tracking of human body parts

Input: K-means image/matrix. Output: Set of blobs

BlobsInitial creation

Page 21: Recognition and tracking of human body parts

Particularly important in human body part recognition.

Can not be fused. Technique to identify skin blobs:

◦ Euclidean distance

BlobsSkin blobs

Page 22: Recognition and tracking of human body parts

Conditions:◦ Blobs have to be neighbors◦ Blobs have to share a large border ratio◦ Blobs have to be of similar color

◦ Small blobs are fused to their largest neighbor Neither of these conditions apply to skin

blobs

BlobsFusion

Page 23: Recognition and tracking of human body parts

Associate blobs to body parts

Body part recognition (I)

Page 24: Recognition and tracking of human body parts

Skin blobs play the key role:◦ Head and Upper body:

Torso identification Face and hands identification

◦ Lower body: Legs and feet identification

Body part recognition (II)

Page 25: Recognition and tracking of human body parts

Body part recognition (III)

Page 26: Recognition and tracking of human body parts

Computational time Background subtraction quality Subject clothing Subject position Number of clusters in K-Means algorithm Skin blobs

Problems (I)

Page 27: Recognition and tracking of human body parts

Problems (II)

Page 28: Recognition and tracking of human body parts

Problems (III)

Page 29: Recognition and tracking of human body parts

Main tasks completed Improvements are required for better

results

Possible future work:◦ Multiple people tracking◦ Detailed body part recognition

Conclusion and future work

Page 30: Recognition and tracking of human body parts

?

Questions, comments