Pattern Recognition for Machine Vision
Lecture notes and assignments applying pattern recognition to machine vision: features, classification, clustering, tracking and gestures.
Teaching machines to see draws on pattern recognition, and this course focuses on applying those techniques to problems of machine vision. It covers image formation and processing, feature extraction, biological object recognition, Bayesian decision theory, classification and clustering. Applications include object detection and recognition, morphable models, tracking and gesture recognition, with a strong hands-on component through experiments and a final project.
What You Will Achieve
- Experiment with color spaces, image segmentation and edge detection
- Interpret ROC curves and use PCA and ICA
- Apply Support Vector Machines and clustering algorithms to image analysis
- Build a simple object tracker by learning video background models
Program Overview
Vision and Features
Getting useful information from images.
- Image formation and processing
- Feature extraction
- Fourier shape descriptors and wavelet transform
Pattern Recognition Methods
Core statistical techniques.
- Bayesian decisions
- Density estimation
- Classification
- Clustering
Applications
Putting the methods to work.
- Object detection and recognition
- Morphable models
- Tracking
- Gesture and action recognition
Course Completion
Job Outlook
These techniques apply to computer vision and machine learning work, including object recognition and tracking.
Good for: Students who want a practical view of classical pattern recognition methods applied to vision problems.
Keep in mind: This 2004 course predates today’s approaches to computer vision, so treat it as foundations. Assignments have no solutions listed, and the required text is Pattern Classification by Duda, Hart and Stork.
Bottom line: A hands-on tour of classic machine vision, useful for understanding the ideas behind newer systems.
Frequently Asked Questions
What kinds of projects did students do?
Past final projects include a system that tracks a fast moving baseball in video, soda can recognition on an iPaq, and a video camera-based mouse device.
What will the assignments have me try?
They include edge detection, template matching, ROC curves, sampling from densities, K-means and EM clustering, and image morphing.
Which textbook is required?
Pattern Classification by Duda, Hart and Stork is required, with Computational Vision by Mallot as optional reading and further books suggested.



There are no reviews yet.