Book Image

Learning OpenCV 5 Computer Vision with Python, Fourth Edition - Fourth Edition

By : Joseph Howse, Joe Minichino
5 (2)
Book Image

Learning OpenCV 5 Computer Vision with Python, Fourth Edition - Fourth Edition

5 (2)
By: Joseph Howse, Joe Minichino

Overview of this book

Computer vision is a rapidly evolving science in the field of artificial intelligence, encompassing diverse use cases and techniques. This book will not only help those who are getting started with computer vision but also experts in the domain. You'll be able to put theory into practice by building apps with OpenCV 5 and Python 3. You'll start by setting up OpenCV 5 with Python 3 on various platforms. Next, you'll learn how to perform basic operations such as reading, writing, manipulating, and displaying images, videos, and camera feeds. From taking you through image processing, video analysis, depth estimation, and segmentation, to helping you gain practice by building a GUI app, this book ensures you'll have opportunities for hands-on activities. You'll tackle two popular challenges: face detection and face recognition. You'll also learn about object classification and machine learning, which will enable you to create and use object detectors and even track moving objects in real time. Later, you'll develop your skills in augmented reality and real-world 3D navigation. Finally, you'll cover ANNs and DNNs, learning how to develop apps for recognizing handwritten digits and classifying a person's gender and age, and you'll deploy your solutions to the Cloud. By the end of this book, you'll have the skills you need to execute real-world computer vision projects.
Table of Contents (12 chapters)
Free Chapter
Learning OpenCV 5 Computer Vision with Python, Fourth Edition: Tackle tools, techniques, and algorithms for computer vision and machine learning
Appendix A: Bending Color Space with the Curves Filter

Improving the 3D tracking algorithm

Essentially, our 3D tracking algorithm combines three approaches:

  1. Find a 6DOF pose with a PnP solver, whose inputs depend on FLANN-based matches of ORB descriptors..
  2. Use a Kalman filter to stabilize the 6DOF tracking result.
  3. If an object was tracked in the previous frame, use a mask to limit the search to the region where the object is now most likely to be found.

Often, commercial solutions for 3D tracking involve additional approaches. We have relied on successfully using a descriptor matcher and a PnP solver for every frame; however, a more complex algorithm may provide some alternatives as fallbacks or as cross-checking mechanisms. This is in case the descriptor matcher and PnP solver miss the object in some frames, or in case they are too computationally expensive to use for every frame. The following alternatives are widely used:

  • Update the previous keypoint matches based on optical flow, and update the previous 6DOF pose based on the homography...