Machine vision refers to the use of computer vision in an industrial application or process where it is necessary to execute a specific function or outcome based on the image analysis done by the vision system.
In this FAQ, we will be covering the typical questions that come up when people start looking into machine vision applications, such as how accurate is it? How quick can it detect something? What are the typical limitations in depth of field? What are the applications and benefits?
Machine vision systems use software to trigger various actions based on the analysis of an image, and are widely accepted for applications, including quality control and automated inspection. Advances in technology have led to more applications being solved using machine vision, including process control and optical sorting - removing undesirable stuff from bulk material along with robotic movement - positioning and orientation of items to be picked up by a robot arm.
How accurate is a Machine Vision System?
This depends on the camera's resolution, the field of view and the size of the object(s) that the system is inspecting or monitoring. We need to account for whether the system is looking at something which is the size of a postage stamp or a large piece of equipment (or something even larger, like a football field), all of which can affect the accuracy of the machine vision system.
For example, if you are using a system on a large football field, the accuracy would be less accurate than if you are looking at an object in a tiny field of view, such as a postage stamp.
An additional consideration is the camera's resolution, how many pixels are there to be able to resolve that image? The camera is looking for contrast changes within the pixels themselves, if a neighbouring pixel changes brightness levels or the intensity changes, this can become a threshold and then you can use interpolation software to increase the accuracy of the identification.
How quickly can it detect an object or a defect?
This comes down to two core items which go hand in hand in terms of capturing and then processing.
- The frame rate of the camera
- The processing capabilities and complexity
The frame rate of the camera refers to how many frames per second it can capture. Once the image is captured, intelligence starts to analyze the image, which is where the system's computer vision comes into play.
The processing capabilities also play a major part and this can depend on the complexity of the system and how it is being used. For example, are you using machine vision to measure an object(s)? Are you checking for defects where it has to compare a captured image to a number of other images that are within its database?
Nowadays, smart cameras are incredibly sophisticated, and IT architectures are becoming more elaborate with the power of capturing, pre-processing and processing all becoming more intelligent.
How close or far can a system see when it is detecting an object?
This is a combination of optics, illumination and the aperture of the camera system, all of which are interchangeable dependent on the type of setup that is required.
Depth of field and focus affect how effectively a camera can image an object, for example, in a large depth of field which is what is in focus and what is not (everything that is in focus is your depth of field).
To get a really large depth of field, this will require significant illumination and a lower aperture (as small as it will go), with most applications, this will give you a large depth of field and this is why lighting/illumination is incredibly important in a number of applications. This also has a lot to do with the resolving power and quality of the optics.
What are the different types of Machine Vision?
There tend to be two types, the first type is a smart camera which is a combination of hardware and software in one “body” within one piece of self-contained hardware. Smart cameras are more prevalent across the market but they do have limited capability as the processing hardware is fixed.
Another type is vision systems that use a “dumb” camera connected to a controller, a PC or industrial PC or a dedicated platform that a company may have created. These types of machine vision systems are more flexible. They are more tuned for the specific functions, high-end applications, and more challenging tasks beyond the capability of a smart camera.
What are the typical applications with Machine Vision?
Quality inspection and control are generally the biggest applications for Machine Vision systems. It enables human operators to spend more time doing manual inspections.
Below is a list of the general applications:-
- 1D/2D Codes
- Track and Trace
- Fill levels
- Tolerance Constraints
- Proximity Detection
- Product positioning
- Robot manipulation
- Vehicle direction
- Detect detection
- Operation confirmation
- Contamination detection
- Monitoring safety
What are benefits of Machine Vision?
There are many benefits including:
- Reducing human interaction and inspection from a process
- Speed – a camera can inspect parts much faster than a human can on average
- Consistency of inspection – compared to human inspection
- Ability to sustain a high level of inspection compared to humans (who, whilst still part of the process, may only be able to do it for a short period of time)
- Automating processes – reducing interaction from humans
- You can fully understand what is going on in your processes to making a particular part and can provide automated data points and millions of data points on your products
- Improve inspections, reduce costs, improve quality and remove the human element of Manufacturing
- Machine vision solutions give customers more flexibility with their operations
- A machine vision camera can read both 1D and 2D codes and do quality verification, part presence and orientation; product defect detection; color inspection; and other visual inspection processes
- Machine vision systems can recognize text (OCR), handwriting (ICR), barcode (OBR) and optical mark (OMR)
- Computer vision, whether used with robots or in quality control systems in manufacturing processes, can process images or video and initiate appropriate actions
- With these capabilities, computer vision systems can conduct a wide range of tasks, well beyond that of barcode reading
- Machine vision can know what is in a standard cardboard box, carton or other containers, so it is good for applications where the packaging has distinguishing characteristics
- It can determine sizes or styles, making it effective in many retail applications