Learning OpenCV 3

Chapter 22. Object Detection

In the previous two chapters, we covered the basics of machine learning and then moved on to investigate, in some depth, a large number of techniques that the OpenCV library provides for discriminative and generative learning. Now it’s time to put it all together, to combine the computer vision techniques we have been learning throughout the book with the machine learning techniques, and to actually apply learning to practical problems in computer vision. One of the most important such problems is object detection—the process of determining whether an image contains some particular object and, where possible, the localization of that object in pixel space. In this chapter, we will look at several methods that achieve these goals, in every case by making use of the lower-level machine learning techniques from the previous chapter.

Tree-Based Object Detection Techniques

Having looked at many of the lower-level methods for machine learning in the library, we now turn to some higher-level functions that make use of those various learning methods in order to detect objects of interest in images. There are currently two such detectors ented on OpenCV. The first is the cascade classifier, which generalizes the very successful algorithm of Viola and Jones [Viola01] for face detection, and the second is the soft cascade, a further evolution of that algorithm that uses a new approach to give what is, in most cases, a more robust classification than the cascade ...