Analysis of HALCON machine vision algorithm package

Recently, I have been using and debugging visual capture technology, and I have a little understanding of it. That's how I came into contact with HALCON.

HALCON is a complete set of standard packages developed by MVc in Germany, with a widely used machine vision integrated development environment. It saves product costs and shortens the software development cycle - HALCON's flexible architecture facilitates the rapid development of machine vision, medical imaging and image analysis applications. It is recognized as the best Machine Vision software in Europe and Japan.

HALCON supports Windows, and Mac OS X operating environments, which guarantees the validity of the investment. The entire function library can be accessed in a variety of common languages ​​such as C, C#, Visual B and Delphi.

HALCON provides interfaces for a large number of image acquisition devices, ensuring independence. It provides interfaces for more than 100 industrial cameras and image acquisition cards, including GenlCam, GigE and IIDC 1394.

Powerful 3D visual processing

A particularly outstanding new technology provided by HALCON 11 is 3D surface comparison, which compares the surface shape measurement of a 3D object with the expected shape. All 3D technologies provided by HALCON, such as multi-view stereo or sheet of light, can be used for surface reconstruction; 3D reconstruction is also supported directly through off-the-shelf 3D hardware scanners.

In addition, photometric stereo methods have been improved for special applications in surfaces. Moreover, HALCON now supports many methods for 3D object processing, such as point cloud calculation and triangulation, feature calculation such as shape and volume, point cloud segmentation through cut planes, etc.

High-speed machine vision experience

Automatic operator parallel processing (AOP) technology is a unique feature of HALCON. HALCON 11 supports more than 75 operators for machine vision algorithms using processing, more than any other software development kit.

In addition, depth image acquisition based on focus change (depth from focus), fast Fourier transform (FFT) and HALCON's local deformation matching have been significantly accelerated. HALCON 11 will bring users a faster machine vision experience.

Let you choose HALCON for the following 3 reasons:

1. In order to allow users to develop visual systems in the shortest possible time, HALCON includes an interactive programming interface HDevelop, in which you can directly write, modify, and execute programs with HALCON, and view all variables in the calculation process. After the design is completed, you can directly output C, C++, VB, C#, vb and other program codes and insert them into your program. HDevelop is also linked to hundreds of example programs. In addition to the description of individual calculation functions, you can also find application examples according to different categories at any time for easy reference. In addition, the problem-oriented manual allows you to find the most appropriate instructions and operating concepts.

2. HALCON does not limit the image acquisition device, you can choose the appropriate device yourself. The manufacturer has provided driver links for more than 60 types of cameras. Even for cameras that are not yet supported, in addition to being able to easily capture images through pointers, you can also use HALCON's open architecture to write your own DLL files and connect to the system.

3. When using HALCON, there are no special restrictions when designing human-machine interfaces, and no special visualization components are required. You can fully use the programming language in the development environment, such as visual studio, .NET, Mono, etc., to build your own interface. The end user cannot see your development tools, and only a very small resource kit is required on the machine that executes the job.

In the industry, halcon is generally used in the lower end of the industry chain. And machine vision is also affected. Halcon used to be the main player in the industry, but now its share is declining. In the past five years, a large number of machine vision manufacturers and startups have emerged in China, while in the past, they were basically imported. These startups used opencv in large quantities in the early stages of their products. However, after a period of iteration, it is no longer just a call. Many commercial companies have combined hardware to optimize and innovate on it, and it is no longer just dependent on calling libraries.

Moreover, the business model of selling software libraries alone is not easy to follow in China, and many are sold together with hardware. In terms of employment, the basic requirement is to be familiar with opencv, halcon is an option, and the use of halcon is often a legacy of project history. If you do other visual directions in the future, people will not use halcon, but opencv is still used quite a lot. Although after you get in, they may have their own set of internal algorithm libraries. But commercial companies cannot require applicants to be familiar with their internal libraries. So they will post opencv. And many internal developments also refer to the architecture of opencv.

As for the problem of the effectiveness of opencv algorithms, in fact, opencv is not specially built for the machine vision industry. So some algorithms are not specifically optimized. The optimization here includes both the optimization of computing speed on a specific processor and the optimization of specific algorithms to solve problems. So what I said before is to emphasize algorithm capabilities. You have to check papers, do experiments and improvements, and even design algorithms yourself to accumulate technology. Instead of treating it as a library and just calling the interface, in other words, you make your own halcon in the enterprise, which actually reflects the improvement of the research and development capabilities of the new generation of Chinese people.

Review editor: Liu Qing