he robot vision system is generally composed of optical systems (light sources, lenses, industrial cameras), image acquisition units, image processing units, execution mechanisms, and human-machine interface modules. The simplest machine vision system includes visual imaging, image processing, and operation control parts.
1. Visual imaging part
Visual imaging also includes several typical components: light source, lens, and industrial camera.
Both light sources and lenses require us to master optical knowledge. Different lighting methods can enable the camera to produce completely different images of objects; The selection of lens magnification, focal length, field of view, etc. directly determines the fidelity of the imaging. For a machine vision engineer, mastering how to choose a lens, how to choose a light source, and how to determine the lighting method are the most basic skills.
Industrial cameras require us to master the knowledge of optoelectronics, the differences between camera sensors, and the basic knowledge of image imaging, such as clarity, dynamic range, field of view angle, etc., so that we can choose the correct camera based on needs and scenes. The fastest way to master this knowledge is to buy an entry-level DSLR to study the relationship between these imaging parameters and imaging.
2. Image processing section
Image processing is generally understood to be carried out on PC machines, but in fact, in the industrial field, most industrial control computers are used because of their stability and cost advantages.
With the development of recent years, embedded hardware is also booming. Many factories can use Open-source hardware such as raspberry pie to achieve small needs such as controlling the switches and status monitoring of hundreds of instrument panels.
For beginners, priority can be given to mastering the development of PC and X86 platforms, which can be extended to embedded platforms after familiarity.
In the software part, most application layers are implemented using C #,. net, QT, and C++, so mastering one of these programming languages is essential; At the level of image algorithms, typical open source algorithms include OpenCV, while commercial ones include Halcon, VisionPro, etc. It is recommended to start with Halcon as a starting point; If you want to further delve into algorithm level, you can study machine learning, which may be the main direction in the future.
In terms of theory, it is more important to master the basic concepts of image processing.
3. Motion control part
A typical motion control card, such as Gugao, can be studied first. A more advanced PLC can also be played, but the difficulty in this part lies in the correction of accuracy, as many scenarios and requirements require high precision.
In addition to the above three points, the construction ability of the overall plan is crucial because the plan needs to connect these parts together and be able to connect with real scenarios to meet the actual production automation needs.
The construction ability of the overall plan depends on a deep understanding of the production process, the connection between all components, and the relationships; Both of these require the accumulation of experience from multiple projects in order to provide a good solution.