17. Image Processing and Data Analysis

Docenti / Lecturers

Alessandro Bevilacqua (UNIBO-DISI)

4 CFU - 32 ORE / 4UEC - 32 HOURS

OBIETTIVI FORMATIVI / LEARNING OBJECTIVES

In the first part of module, the student learns the fundamentals of a computer vision system, made of a camera and image processing and analysis software, to be used for metrology and quantitative analyses. The student acquires the basic principles of optics and radiometry useful to understand the image formation on the sensing device and the basics of image processing operations and image analysis. The student knows the principles of 3D imaging. Some commercial and open source software for image processing and analysis will be presented.   

Therefore, the high-level learning objectives are:

  • Basic knowledge of optics, radiometry and image sensing
  • Understanding the process of image formation
  • Expertise in image processing – point, local and global operations
  • Capability to design a simple computer vision system
  • Capability to extract radiometric and geometric measures from images
  • Understanding the principle of 3D imaging

In the second part of module, the student will learn the fundamentals of pattern recognition and machine learning, that is, how to automatically extract meaningful information from images and dataset. The student acquires the expertise to accomplish this task using either parametric or nonparametric models, in a supervised or unsupervised manner. The student knows how to setup and validate a model, learning the concepts of overfitting and generalizability. Useful examples exploiting 1D signals, images and multi-feature data will be employed.

Therefore, the high-level learning objectives are:

  • Principles of pattern recognition and machine learning
  • Understanding descriptive statistics and statistical inference
  • Expertise in parametric and nonparametric regression
  • Capability to explore data relations to extract meaningful features
  • Knowledge of supervised and unsupervised learning
  • Understanding model validation