Understanding how machine vision works will help you determine if machine vision solves specific application problems in manufacturing or processing.
People often do not understand what machine (computer, artificial) vision can and cannot do for a production line or process. Understanding how it works can help people decide if it will solve problems in a given application. So what is computer vision and how does it actually work?
Artificial vision is a modern technology that includes tools for taking, processing and analyzing images of the physical world in order to create information that can be interpreted and used by a machine through digital processes.
Application of artificial vision in industry
Computer vision refers to the use of one or more cameras to automatically inspect and analyze objects, most commonly in an industrial or manufacturing environment. The resulting data can then be used to manage processes or production activities.
This technology allows you to automate a wide range of tasks, providing machines with the information they need to make the right decisions in each of the tasks assigned to them.
The use of artificial vision in industry allows the automation of production processes, which leads to better production results through the application of quality control and greater flexibility at each stage.
Nowadays, the use of industrial artificial vision has made it possible to significantly improve production processes. This has made it possible to obtain higher quality products at lower costs in almost all areas of industry, from automotive and food, to electronics and logistics.
A typical use would be in an assembly line where a camera is started after an operation is performed on a part that takes and processes an image. The camera can be programmed to check for the position of a particular object, its color, size or shape, and for the presence of an object.
Machine vision can also search and decode standard 2D matrix barcodes, or even read printed characters. After inspecting a product, a signal is usually generated that determines what to do next with the product. The part can be dropped into a container, sent to a fork conveyor, or passed on to other assembly operations, and the inspection results are tracked in the system.
In any case, computer vision systems can provide much more information about an object than simple position sensors.
Computer vision is commonly used for, for example:
- providing quality assurance,
- robot (machine) control,
- testing and calibration,
- real-time process control,
- data collection,
- machine monitoring,
- sorting and counting.
Many manufacturers use automated computer vision instead of inspection personnel because it is more suitable for repetitive inspections. It is faster, more objective and works around the clock.
Computer vision systems can inspect hundreds or thousands of parts per minute and provide more consistent and reliable inspection results than humans. By reducing defects, increasing revenue, making compliance easier, and tracking parts with computer vision, manufacturers can save money and increase their profitability.
How machine vision works
The discrete photocell is one of the simplest sensors in the field of industrial automation. The reason we call it “discrete” or digital is because it only has two states: on or off.
The principle of operation of a discrete photocell (optical sensor) is to transmit a light beam and determine whether the light is reflected from the object. If there is no object, no light is reflected into the photocell receiver. An electrical signal is connected to the receiver, usually 24 V.
If the object is present, the signal is turned on and can be used in the control system to perform an action. When the object is removed, the signal is turned off again.
Such a sensor may also be analog. Instead of two states, i.e. off and on, it can return a value indicating how much light is returned to its receiver. It can return 256 values, from 0 (meaning no light) to 255 (meaning a lot of light).
Imagine thousands of tiny analog photocells arranged in a square or rectangular field pointed at an object. This will create a black and white image of the object based on the reflectivity of where the sensor is pointing at. The individual scan points in these images are called “pixels”.
Of course, thousands of tiny photoelectric sensors are not used to create an image. Instead, the lens focuses the image on a semiconductor array of light receivers.
This matrix uses arrays of photosensitive semiconductor devices such as CCD (Charge Coupled Device) or CMOS (Complementary Metal-Oxide-Semiconductor). The individual sensors in this matrix are pixels.
Four main components of a computer vision system
The four main components of a computer vision system are:
- lenses and lighting ;
- image sensor or camera ;
- processor ;
- a way to transfer results, whether it be a physical input/output (I/O) connection or another method of communication.
Computer vision can use color pixel scanning and often uses a much larger array of pixels. Software tools are applied to captured images to determine the dimensions, edge placement, movement, and relative position of elements relative to each other.
The machine vision lenses capture the image and transmit it to the sensor in the form of light. To optimize a computer vision system, the camera must be paired with the correct lenses.
Although there are many types of machine vision lenses, fixed focal length lenses are commonly used in computer vision applications. When choosing, three factors are important: field of view, working distance, camera sensor size.
Lighting can be applied to an image in a variety of ways. The direction in which the light arrives, its brightness, and its color or wavelength compared to the color of the target are very important factors to consider when designing a computer vision environment.
While lighting is an important part of getting a good image, there are two other factors that affect how much light an image receives. The machine vision lens includes a setting called an aperture that opens or closes to allow more or less light to enter the lens.
Combined with the exposure time, this determines the amount of light that hits the pixel array before lighting is applied at all. The shutter speed or exposure time determines how long the image is projected onto the pixel array.
In computer vision, the shutter is electronically controlled, usually with millisecond precision. After capturing the image, software tools are applied. Some are used before analysis (pre-processing), others are used to determine the properties of the object under study.
In the pre-processing stage, you can apply effects to an image to sharpen edges, increase contrast, or fill gaps. The purpose of these tasks is to expand the capabilities of other software tools.
Purpose of computer vision
Here are some common tools you can use to get information about your goal:
- Number of Pixels: Specifies the number of light or dark pixels in the object.
- Edge detection: finds the edge of an object.
- Measurement (metrology): measurement of the dimensions of an object (for example, in millimeters).
- Pattern recognition or pattern matching: Finding, matching or counting specific patterns. This may include detecting an object that may be rotated, partially obscured by another object, or have other objects.
- Optical Character Recognition (OCR): Automatic reading of texts such as serial numbers.
- Barcode reading, data matrix and “2D barcode”: collection of data contained in various barcoding standards.
- Blot Detection: Checks an image for spots of interconnected pixels (such as a black hole in a gray object) as a guide to the image.
- Color analysis: identification of parts, products and objects by color, quality assessment and selection of elements by color.
The purpose of extracting data from checks is often to use it for comparison against target values to determine a pass/fail or pass/fail result.
For example, when checking a code or barcode, the received value is compared with the stored target value. In the case of measurement, the measured value is compared with the correct values and tolerances.
When checking an alphanumeric code, the OCR text value is compared to the correct or target value. To check for surface defects, the measured defect size can be compared to the maximum size allowed by quality standards.
Machine vision in industry has huge potential. These artificial vision systems, applied in robotics, make it possible to offer an automatic solution at various stages of production, such as quality control or the detection of defective products.
Quality control is a set of techniques and tools. It will allow us to identify errors in the production process, as well as take appropriate measures to eliminate them. This provides much more control over the final product, which ensures that when it reaches the consumer, it will meet specific and established quality standards.
Thus, products that do not meet the minimum quality requirements are excluded from the process, which allows to eliminate possible failures that occur in the production process. This is achieved by conducting inspections and random tests on an ongoing basis.
The use of quality control in production has several advantages:
- Performance improvement ;
- Reduction of material losses ;
- Price reduction ;
- The best quality of the final product.
Communication in computer vision
Once received by the processor and software, this information can be communicated to the control system via a variety of industry standard communication protocols.
Major computer vision systems often support EtherNet/IP, Profinet and Modbus TCP protocols. The serial protocols RS232 and RS485 are also common.
Digital I/O is often built into systems to run and simplify reporting of results. Computer vision communication standards are also available.
Artificial vision systems have a wide variety of applications. It can be adapted to different industries and the different needs of each production line. Today, any company that manufactures products to a certain standard can benefit from the benefits of computer vision as part of their manufacturing process.
Understanding the physical principles and capabilities of artificial vision systems can be helpful in determining whether such a technology is suitable in a manufacturing process in a given case. In general, everything that the human eye sees, the camera can see (sometimes more, sometimes less), but decoding and transmitting this information can be quite difficult.