Pattern Classification

Scott E. Umbaugh · CRC Press eBooks · 2022

Pattern classification is a critical process for computer vision applications and is often the final step in algorithm development. With these applications, the goal is to identify objects in order for the computer to perform a vision-related task. The types of computer vision tasks to which these methods are applied are many and varied, ranging from computer diagnosis of medical images to object classification for robotic control. In this chapter, we define pertinent terms, conceptually discuss the most widely used current methods and look in detail at some of the basic classification algorithms. Training and testing methods are discussed, including cross-validation techniques. Categories of classification algorithms, along with examples of each are discussed including basic methods nearest neighbor (NN), K-NN, nearest centroid and template matching. More advanced methods including Bayesian, support vector machines and random forest classifiers are introduced. Newer advances in deep learning, including convolution neural networks, are discussed. Cost and risk functions for success measurements are included. Tools for pattern classification including Python, R: Bayesian modeling, Matlab and CVIPtools development environments are discussed. Along with the text are illustrative figures and associated monochrome and color images. The end of chapter exercises includes problems and programming exercises.

Cited by 29 works. doi:10.1201/9781003221135-7

Authors