MACHINE VISION SYSTEMS DESIGN

Machine vision systems are now rapidly losing their former reputation as complex, unreliable and expensive technology. Developments in the field of image sensors and especially the performance of today’s processors have brought the possibilities of machine vision closer to the requirements of users. The number of applications is growing so fast. The growth in the number of deployments is to a large extent also caused by the growing demands on the so-called total quality of production, which can only be achieved by inspecting each piece produced. The human factor must be excluded from such control, because it is true that “no one is perfect”. Often, machine vision is not only the best, but also the only possible way to implement the required control.

The overall organization of the task

By choosing the spatial arrangement, we already determine to a certain extent the type and number of cameras and lighting units and the requirements for the focal lengths of the machine vision lenses. We must be clear from which direction and distance the camera will capture the scene, while we must determine the sufficient resolution of the camera, or the need to use multiple cameras. At the same time, we have to decide on the method of lighting and its color – this is also related to the elimination of disturbing light and thus to the design of shading and the possible need to use color or polarizing filters in the camera.

Technical and software means of image data processing

Here we have a choice between two concepts:

The use of so-called smart cameras and image processing within the capabilities of these cameras
Connecting cameras to a computer and processing image data with a standard computer
Surprisingly, the price of the final solution is not an essential criterion here – both concepts are approximately equally expensive. The requirement for computing power, flexibility and variability of software is especially important. Smart cameras themselves process image data and are usually equipped with binary outputs, allowing to signal the result of the process. They usually do not allow free programming, they can only be configured via a serial line or Ethernet connection. They are usually equipped with specialized signal processors or low-power RISC processors with a clock speed in the order of hundreds of MHz and simple real-time operating systems. These facts already point to their limitations. Smart cameras are equipped with only a few basic image processing tools and are only suitable for simple tasks. On the other hand, a large number of tasks are usually solved by surprisingly simple means, and here these integrated cameras will work. Estimating the situation in advance requires a lot of knowledge, emotion and experience. For example, when it is necessary to deal with a changing scene, respond to changes in the position, number and shape of objects, changes in lighting or solve complex and performance-intensive algorithms, we quickly encounter limits that are fixed and insurmountable. The effort to solve tasks beyond their capabilities with smart cameras is behind many unsuccessful projects. On the other hand, a large number of tasks are usually solved by surprisingly simple means, and here these integrated cameras will work. Estimating the situation in advance requires a lot of knowledge, emotion and experience. For example, when it is necessary to deal with a changing scene, respond to changes in the position, number and shape of objects, changes in lighting or solve complex and performance-intensive algorithms, we quickly encounter limits that are fixed and insurmountable. The effort to solve tasks beyond their capabilities with smart cameras is behind many unsuccessful projects. On the other hand, a large number of tasks are usually solved by surprisingly simple means, and here these integrated cameras will work. Estimating the situation in advance requires a lot of knowledge, emotion and experience. For example, when it is necessary to deal with a changing scene, respond to changes in the position, number and shape of objects, changes in lighting or solve complex and performance-intensive algorithms, we quickly encounter limits that are fixed and insurmountable. The effort to solve tasks beyond their capabilities with smart cameras is behind many unsuccessful projects. number and shape of objects, changes in lighting or solve complex and performance-intensive algorithms, we quickly encounter limits that are fixed and insurmountable. The effort to solve tasks beyond their capabilities with smart cameras is behind many unsuccessful projects. number and shape of objects, changes in lighting or solve complex and performance-intensive algorithms, we quickly encounter limits that are fixed and insurmountable. The effort to solve tasks beyond their capabilities with smart cameras is behind many unsuccessful projects.

Connecting cameras to a standard computer is a necessary choice for more complex applications, but even for simpler applications it leaves us more room to correct any initial inaccuracies in the requirements estimate. The performance of modern processors dramatically exceeds even the best smart cameras, and the built-in computer no longer has to look like a large box with several fans. In addition, many typical image data operations can be accelerated by parallel processing on multiple cores simultaneously. Some software systems, and the pioneer of these technologies, such as the VisionLab machine vision system, are able to take advantage of the powerful massively parallel performance of today’s graphics processors. While today’s CPUs have up to four cores, the GPU divides the calculation into 240 cores, for example.

Even in the best programming environment, we can sometimes encounter the absence of the required function or its insufficient performance. The open possibility of adding your own code can give us peace of mind. And if that doesn’t help meet our requirements, at worst, we can change the entire machine vision software system to avoid order failure. Larger room for maneuver is always appropriate.

In many cases, it is sufficient if the only output of the visual inspection system is a binary output, signaling a defective product. Increasingly, machine vision systems no longer operate separately from the rest of the world, but their integration into the company’s information systems is required. The software environment should enable the inclusion of visual inspection in the broader context of the control and visualization system, it should transmit all, including video, data in the network, communicate with PLC, cooperate with SQL servers, HTTP servers, etc.

Next, suppose we design a system where the cameras are connected to a standard computer. It remains to choose the appropriate camera, machine vision lens and lighting.

Camera

There are many criteria for selecting a camera, we can consider a CCD or CMOS detector, chip size and resolution, black and white or color sensor. In the case of a color sensor, it can be a single-chip design, a three-chip design or sequential scanning with a black-and-white sensor and color filters. The camera connection can be analog or digital. In the case of a digital interface, we have a choice of Ethernet / IP, USB or Firewire. Digital cameras can either run at a fixed frame rate, can be externally triggered, can run freely with light accumulation, can provide differently compressed data streams, or can provide undistorted raw data, etc. – A really confusing number of criteria play a role in camera selection.

Read more:
https://www.timemagazine.org/four-main-components-of-machine-vision-system/
https://www.bestmag.org/machine-vision-principles-and-characteristics/
https://www.postingstation.com/machine-vision-system-hardware-the-4-basics/
https://www.articlering.com/machine-vision-several-pitfalls-of-systems-design/
https://www.nytoday.org/machine-vision-principle-application-and-development/
https://shop.heyluu.com/531-machine-vision-lenses
https://magnifiedads.com/market/machine-vision.html
https://flemicon.com/understand-the-lenses-for-machine-vision
https://thewion.com/development-of-food-safety-by-machine-vision/
https://www.folkspaper.com/topic/machine-vision-sensor-resolution-and-size-5650239245516800.html