Artificial vision - camera control systems; consists of different disciplines such as electronics, optics, photonics, software and automation. You need to master all these technologies to develop a successful image processing solution. Today, camera and sensor technology is developing rapidly. Processor powers are increasing, powerful graphics processors are multiplying and cheaper, different camera types are entering the market. Artificial intelligence software is the breakthrough in image processing.

As AKU Vision Technologies, we offer solutions in almost every aspect of image processing. By following current trends, we are able to analyze new products entering the market and offer our original solutions and contributions.

(A typical vision system)

Our solutions generally have the following capabilities because they are mostly industrial systems.

• Producing 5-24V electrical signal (for the purpose of line stop, alarming, lighting etc. depending on OK or NOK)
• Saving all processed images with results, retrospective review, writing to the database, statistics and reporting
• Communication with automation systems such as PLC, conveyor, motor, etc. (IO, MODBUS, ProfiBUS etc. communication)
• Having convenient end-user screens and interfaces that match the user's corporate identity
• 100% integration for Industry 4.0, IıoT, and BigData solutions

 

As AKU Vision Technologies, we mainly use computer vision systems in our solutions. We design systems where one or more cameras are connected to a powerful computer, the main job being software. Altough less in number, we also have smart camera solutions. Intelligent camera systems or computer based systems are artificial vision techniques that have different abilities and objectives that serve essentially the same purpose.(Mentioned below)
In order to look more deeply into artificial vision technology, let's briefly discuss the basic concepts used in image processing.

1. Camera: Devices that we use to capture images. Although its generally thought to be a camera+lens system, the camera is independent of the lens in the artificial vision world. When we say camera, we mean a device which usually a lens can be externally attached, that has a sensor in it, ans so can transfer images. Terminology is named with its resolution, sensor characteristics and communication protocol.

3 MegaPixel, Global Shutter, USB3 Camera or 5 MegaPixel, Rolling Shutter, Gigabit Ethernet Camera.

There are many different types of cameras in the market that are customized for different purposes such as Spectral, LWIR, SWIR, Infrared, Thermal etc. In our Blog and Market pages, more detailed technical information about different cameras and their application areas will be given.

 

2. Lens: Lenses as we all know and use in photography. Lenses attached to the cameras are classified according to their working modes (telecentric, pericentric, zoom, normal etc.) according to the types of focus (auto-focus, manual focus, variable focus etc.).

There are many details in lens selection, but the most important one is the information like the suitability for the camera which it will be attached (mount point), it's resolution (yes the lens has a resolution too), the focus distance, and the way it works. A notation like “5MegaPixel, C-Mount, 25 mm, Fixed Focus lens” is used when named.

Images taken with 25 and 6 mm lens. The larger the value of the lens mm, the narrower the angle. Narrow angled lenses give better results against perspective deterioration as seen in the above image. Telecentric lenses should be used to prevent perspective distortion.

 

3. Sensor: The electronic section in the camera that actually gets the image and determines all the properties of the image taken, including the quality. Unlike camera manufacturers, there are not so many sensor manufacturers in the market. Since camera manufacturers work with different sensor manufacturers, it is more professional to specify the sensor information when talking about the camera.

3 MegaPixel, USB3 Camera with Sony IMX 265 Sensor

 

4. Communication Protocol: There should be a method of communication for such things as transferring the image taken by the cameras and establishing a connection with the computer. Among the different industrial protocols, the most commons are USB and Ethernet communications (as can be guessed). Other than those there are many other forms of connection such as CameraLink and CoaxPress. (We will talk more about this at our blog pages)

USB3.1 (USB3Vision in artificial terminology) is the most preferred USB with its numorious advantages like high bandwidth and compatibility

 

Ethernet (Gigabit Ethernet) or as used in artificial vision terminology GigEVision is used for cases where the USB distance is not enough and data is being sent at a longer distance.

Bandwidth and speed have become much more important in recent years, with the increase in resolution and frame per second (FPS) values and number in multicamera applications. (Referred to in future articles)

Once we have made these basic definitions, we can classify our image processing applications.

Area Scan Image Processing Systems:

The most commonly used artificial vision systems are those that use a area scan sensor. Here, the camera takes a picture of a specific area, just as it is in a typical picturing machine. The photo is expressed in terms of resolution by the size of the sensor according to its type. (Such as 640 x 480)

One of the most important points for field scan cameras is how the sensor works. As a general approach, it is logical to use a Global Shutter sensor if the image is moving and the Rolling shutter if it is stationary.

Global shutter sensors must be used in situations where the object is moving (turning, moving, etc.) or if vibration occurs in the environment. Global shutter is a more complex technology and is generally more expensive than Rolling shutter sensors. However, "Rolling Shutter" sensors also have some advantages like being more economical and having less noise. If a fixed object is photographed, the Rolling shutter makes more sense (eg for a measurement application )

Line Scan Image Processing Systems:

The logic here is similar to that of a desktop scanner or copy machine. The main difference is that the line sensor is in motion in the scanner or copier,but here the camera is stationary and the object to be photographed must move. The camera takes a single line image. When talking about the camera resolution, it is specified how many pixels this single line is. Generally 2K, 4K, 8K notation is used. A 2K line scan camera indicates that the only 1 line has 2048 pixels laterally. We can get a very high resolution picture by taking as many consecutive lines as we want.

When we receive 8192 rows in succession with a 2K linescan camera we get a 16 MPixel image. Such cameras are often used in situations such as photographing moving objects on a conveyor or inspecting a fast flowing fabric. In general, the encoder is connected to the camera so that the increase or decrease in speed of the conveyor or the object does not affect the image received. They are used in areas such as fabric and print inspection and precise measurement of fast moving parts.

A typical linescan camera. Notice that the sensor contains a single line.

3D Cameras:

They are the cameras that can give the 3. dimension (z). Stereovision (two cameras that are calibrated relative to each other, just like the human eyes) is commonly used. Recently, ToF (Time of Flight) cameras have also become widespread. altough being more limited in usage, there are also Laser Assisted Line Profile cameras. Just common to all is the ability to take photographs where each pixel's distance (z) in space is given which we call point cloud or disparity map.

It is used in applications such as robot eye (robotic pick & place), height volume detection and processing of 3D objects.

Computer Vision Systems and Smart Cameras

A camera is called a smart camera if it houses a smart processor that can be programmed and configured for various tasks. If the camera does not have such a processor unit, it only transfers the image it receives and the transferred image is processed by a computer software, this is called a computer vision system. These 2 technologies are two competing approaches, both in terms of producers and practitioners. There are places where both are advantageous and disadvantageous. In general, if the problem to solve is simple and well-defined, smart cameras; if they are complex, variable and more professional computer-based systems are preferred. Smart cameras can be learned by going through a shorter training, by changing a variety of parameters a fast-paced application can be designed.. On the computer side, a basic mastery of software knowledge, cameras, lenses, image processing libraries and artificial vision technology will be required.

Smart cameras are usually slow-acting , oftenly with integrated lighting and electrical input / output ports, and which can easily be distinguished easily by their larger cases. Smart camera manufacturers are companies that mainly produce automation sensors, scanners, components such as Cognex, Keyence, Sick, Omron, Panasonic, Banner, Leuze.

In computer vision systems on the other side, it is possible to encounter almost any kind of camera and hardware. The software or libraries that run inside the computer are as determinant as the hardware.

Lighting :

In artificial vision practice, the most important factor to be emphasized is the selection of an accurate illumination. In our blog and market pages, you will find detailed information about many different lighting types. As the topic is detailed, we will provide information about different technologies in the future. As AKU Vision, we offer a wide range of lighting products, including our own lighting modules.

Current Trends:

As AKU, we keep up with current trends in image processing and adapt it to our solutions. We examine Spectral Imaging, new generation cameras like SWIR or LWIR cameras , new lenses such as Pericentric lenses, UV lighting, LED lighting with high CRI value, and use them in our solutions. In addition , we adapt deep learning and machine learning technologies, and present them to our customers. We follow leading image processing software such as Google Vision API, IBM Watson, and NVidia cuBlas.

With SWIR cameras working outside the visible spectrum of the light, it is possible to capture many details that are not visible. Above,normal and SWIR camera image in foggy environment.

Displaying different liquids (acid, alcohol, water, etc.) with the SWIR camera Note that different liquids with chemically different contents are displayed completely differently with the SWIR camera.

(The SWIR camera will need the appropriate nanometer-powered lighting elements for successful operation.)

We provide the most up-to-date technology to our customers both as hardware and software. With our trainings, we share our technology with our customers in our place or on demand.

In summary AKU; is not a distributor of a specific brand or product but is an image processing company trying to adapt the ideal technology in order to offer the best solution to its customers.

Machine vision, image processing, current trends, technological issues and new products will be featured on our blog and market pages.