Machine vision facts for kids

An early machine vision system from 1983. It uses a camera to see objects on a light table. The image is then analyzed by a computer.

Machine vision (often called MV) is a cool technology. It uses cameras and computers to "see" and understand things automatically. Think of it like giving robots or machines eyes and a brain!

This technology helps machines inspect products, control processes, and guide robots. It's mostly used in factories and industries. Machine vision combines many different technologies. This includes special cameras, software, and hardware. It's a bit different from computer vision, which is more about the science of how computers can see. Machine vision focuses on solving real-world problems.

The whole process starts with planning what the machine needs to "see." Then, a solution is built. When it's running, the system takes pictures. After that, it automatically analyzes these pictures. Finally, it pulls out the information it needs.

What is Machine Vision?

Machine vision is all about getting information from images automatically. It's not just about making a new image. Instead, it takes a picture and figures things out from it. For example, it can tell if a part is good or bad. It can also find out where an object is and how it's turned.

This technology is used for many things. It helps with automatic checks in factories. It also guides robots and helps control how things are made. You might even find it in security cameras or self-driving cars.

Machine vision uses lots of different tools. This includes special software and hardware. It also needs smart people to set it up. In factories, "machine vision" is the main term used. In other areas, like security, other terms might be used.

Machine vision is a type of systems engineering. This means it puts existing technologies together in new ways. It solves real-world problems, especially in factories. It helps machines work smarter and faster.

How Machines "See" and Sort Things

One main use for machine vision is to check and sort items automatically. It also helps guide robots. Let's look at how this automatic checking works.

First, engineers plan exactly what the system needs to do. Then, they build the solution. Here's what happens when the system is actually working:

How the System Works

The first step is to take a picture. This is usually done with special cameras, lenses, and lights. These are chosen to make the important parts of the object stand out.

Then, machine vision software takes over. It uses different ways to process the digital image. This helps it find the information it needs. Often, it then makes decisions. For example, it might decide if a product "passes" or "fails" a quality check.

What Equipment is Used?

An automatic inspection system usually has a few key parts:

• Special lighting to make objects clear.
• A camera or other device to take the picture.
• A processor, which is like the system's brain.
• Software that tells the processor what to do.
• Output devices that show results or control other machines.

Taking Pictures (Imaging)

The camera can be separate from the main computer. Or, it can be built right into it. When they are together, it's often called a "smart camera." This means the camera can do its own processing.

Sometimes, the camera connects to a special computer part called a frame grabber. This helps it send images quickly. Other times, digital cameras can connect directly to a computer. They use common connections like USB or Gigabit Ethernet.

Most machine vision uses regular 2D cameras that see visible light. But there are other options too:

• Multispectral imaging: Sees light in many different colors.
• Hyperspectral imaging: Sees even more colors, like a super detailed rainbow.
• Infrared imaging: Sees heat, which is useful for some materials.
• Line scan imaging: Takes pictures one line at a time, good for moving things.
• 3D imaging: Creates a 3D model of an object.
• X-ray imaging: Sees inside objects.

For 2D images, systems can see in black and white (monochromatic) or color. They also vary in how fast they take pictures (frame rate) and how clear the pictures are (resolution).

While most systems use 2D images, 3D imaging is growing. One common way to get 3D images is called "triangulation." A laser shines a line onto an object. A camera sees this line from a different angle. How the line bends tells the system about the object's shape. Many lines are put together to create a 3D map.

Another 3D method is "time of flight." This sends out light and measures how long it takes to bounce back. This tells you how far away things are.

Processing the Images

After a picture is taken, it needs to be processed. This "thinking" is done by special computer chips. These can be a CPU, a GPU, or an FPGA. Sometimes, they use a mix of these.

Newer methods like deep learning need even more processing power. The image is usually processed in several steps. It ends with the desired result.

Here are some common ways machine vision processes images:

• Stitching: Combines several images to make one bigger picture.
• Filtering: Changes the image to make it clearer or highlight certain features.
• Thresholding: Turns parts of the image black and white. It uses a specific brightness level to decide. This helps separate objects from the background.
• Pixel counting: Counts how many light or dark dots (pixels) are in an area.
• Segmentation: Divides the image into different parts. This makes it easier to analyze each part.
• Edge detection: Finds the outlines of objects.
• Color Analysis: Uses color to identify parts or check their quality.
• Blob detection: Finds groups of connected pixels, like a dark spot on a light surface.
• Neural networks / Deep learning: These are like smart brains that learn from many examples. They can identify objects or classify them as "pass" or "fail." This technology is making machine vision much more powerful.
• Pattern recognition: Finds specific shapes or patterns. This can include finding an object even if it's turned or partly hidden.
• Barcode and 2D barcode reading: Reads codes like the ones you see on products.
• Optical character recognition (OCR): Reads text, like serial numbers.
• Gauging/Metrology: Measures the size of objects very precisely.
• Comparison: Checks if the information found matches what it should be. For example, it checks if a barcode is correct. Or, it checks if a measurement is within the allowed size.

What Comes Out? (Outputs)

A common result from these systems is a "pass" or "fail" decision. If an item fails, the system might trigger a machine to remove it. Or, it might sound an alarm.

Other outputs include:

• Information about an object's position and how it's turned. This helps guide robots.
• Exact measurements of objects.
• Data read from codes or characters.
• Counts of objects.
• Displays showing what the system is seeing or its results.
• Stored images for later review.
• Alarms from security systems.
• Signals to control other machines in a factory.
• User interfaces for people to interact with the system.

How Machine Vision Guides Robots

Machine vision often tells a robot exactly where an object is. It also tells the robot how the object is turned. This helps the robot grab the product correctly. This technology can also guide simpler movements, like a machine moving along one or two paths.

Just like with inspection, the process starts with planning. Then, a solution is built. The technical steps are similar to automatic inspection. But here, the main goal is to give the robot precise location and turning information.

See also

In Spanish: Visión de máquina para niños