What Knowledge Do You Need to Master to Learn Machine Vision?

If you want to learn about machine vision if you never know it before, you may wonder: what is machine vision? What knowledge must be mastered if you want to learn machine vision?

Firtly, let me explain about what is "machine vision".

Machine vision is a rapidly developing branch of artificial intelligence. In short, machine vision is to use machines to replace human eyes to measure and judge. The machine vision system converts the captured object into image signal through machine vision products (i.e. image pickup device, divided into CMOS and CCD), and transmits it to the special image processing system to obtain the shape information of the captured object, which is transformed into digital signal according to pixel distribution, brightness, color and other information; The image system performs various operations on these signals to extract the characteristics of the target, such as area, quantity, position and length, and then controls the on-site equipment action according to the discrimination results to realize automatic recognition according to the preset permissibility and other conditions, including size, output result, angle, quantity, pass / fail, existence / non existence, etc.

And what knowledge do you need to master to learn machine vision. An simple site of machine vision system consists of three parts: visual imaging part, image processing part and motion control part.

1.  Visual imaging part

Visual imaging includes several typical components: light source, lens and industrial camera.

Both light source and lens need us to master optical knowledge. Different lighting methods can make the camera produce completely different images of objects; The choice of lens magnification, focal length and field of view directly determines the fidelity of imaging. For a machine vision engineer, mastering how to select the lens, how to select the light source and how to determine the lighting method are the most basic skills.

Industrial cameras require us to master the photoelectric knowledge, the differences between camera sensors, and the basic knowledge of image imaging, such as definition, dynamic range, field angle, etc., so that we can select the correct camera according to the demand and scene. The fastest way to master this knowledge is to buy an entry-level SLR to study the relationship between these imaging parameters and imaging.

2.  Image processing part

We generally understand that image processing is carried out on PC machines. In fact, in the industrial field, most industrial computers are used because of their stability and cost advantages.

For beginners, they can give priority to mastering the development of PC platform and X86 platform, and can be extended to embedded platform after they are familiar with it.

In the software part, most application layers adopt C # Net, QT and C + +, so it is necessary to master one of the programming languages; At the image algorithm level, typical open source algorithms include opencv, and commercial ones include Halcon, visionpro, etc. it is suggested to start with Halcon; If you want to go further at the algorithm level, you can study machine learning, which may be the main direction in the future.

In terms of theory, it is more about mastering the basic concepts of image processing. This book Machine Vision Algorithms and Applications is worth reading.

3. Motion control part

A typical motion control card, such as Gugao, can be studied. More advanced PLC can also be played. The difficulty in this part is the correction of accuracy, because many scenes and requirements have very high requirements for accuracy.

In addition to the above points, the construction ability of the overall scheme is very important, because the scheme needs to connect these parts and can be connected with the real scene to meet the actual production automation needs.

The construction ability of the overall scheme depends on the in-depth understanding of the production process and the connection and relationship between all components. These two points need the experience accumulation of multiple projects to give a good solution.