Vision algorithms by the book

E-mail

Many engineers who work with machine-vision software might not care to know how algorithms process images to produce useful inspection data. Some engineers, though, like to know how these algorithms work, or they may need to understand algorithms so they can write custom vision software.

If you want to learn more about how vision software works “under the hood,” you may find value in the new book Machine Vision Algorithms and Applications (Ref. 1), which covers machine vision in three main chapters. The longest chapter covers vision algorithms, while the remaining two address vision applications and hardware (lenses, cameras, and lights). If you lack experience working with vision software, peruse an elementary machine-vision book before you tackle this one.

Although the authors claim readers need little knowledge of mathematics to use the book, a review of calculus, sets, and Boolean algebra might help. In many cases, though, the authors simply present equations and do not expand upon them in the text. But even without worked-out math examples, the algorithm explanations will help you better understand how vision operations find edges, measure circle parameters, match templates, and so on.

This book has an academic tone—Wiley sells it as a textbook—so in places it sounds pedantic. The conversion from German has some rough spots: The translators use the word discretize, for example, in place of digitize. (And, some of the references cite German publications, which can be difficult to find in North America.)

I found the material on camera calibration particularly useful because the authors thoroughly discuss sources of measurement errors and how to adjust for or correct them. Again, readers get a lot of math equations, but the figures and text by themselves provide useful information. Although commercial vision software usually includes some calibration capabilities, engineers and designers can benefit from a better understanding of error sources and calibration techniques.

The book’s applications chapter offers software examples that work with the free student version of Halcon 8.0, a vision-software package from MVTec Software. The book’s preface provides the download Web address and registration details. After registering, you can download the Halcon software and the book’s examples and test images. Stand-alone files give you access to many documents—from a basic introduction to reference and programming manuals.

Access to these programs and files is worth the price of the book. You can create only one student license per book, but this license lasts for two years. The Halcon software runs under the Windows, Linux, and Solaris operating systems.