1. Introduction to industrial cameras
Industrial cameras differ from the cameras on our smartphones or DSLR cameras. They can operate in harsh environments such as high temperatures, high pressures and dusty conditions. Industrial cameras mainly consist of array cameras and line scan cameras.
Linear scan cameras are mainly used in scenarios that require high precision and fast movement, while array cameras have a wider range of applications.
Line Scan Cameras
These cameras have a linear configuration and are typically used in two scenarios. First, they are used to inspect elongated, belt-like fields of view, usually on rotating drums. Second, they are chosen for applications that require a wide field of view with high precision. The two-dimensional images we see in linear scan cameras are made up of multiple linear scans.
Advantages of line scan cameras include the ability to have a large number of one-dimensional pixels, fewer total pixels compared to matrix cameras, flexible pixel sizes, and high frame rates. This makes them particularly suitable for measuring one-dimensional dynamic targets.
Matrix cameras
Array cameras are most widely used in machine vision applications. The advantage of CCD array cameras is their ability to directly capture two-dimensional image information, providing intuitive measurement images.
They can be used for short exposure times, which is beneficial for capturing dynamic scenes, and are also suitable for static objects. Since I mainly use array cameras, this section will focus on array camera selection.
2. Selection of industrial cameras
(1) CCD/CMOS
For static subjects, CMOS cameras are an economical option. However, for moving targets, CCD cameras are preferable. If high-speed acquisition is required – referring to the speed of collection, not the speed of movement – CMOS cameras, with their superior collection rates, should be considered. For high-quality imaging, such as size measurement, CCDs are recommended as they generally outperform CMOS in small sensors.
Industrial CCD cameras are primarily used to capture images of moving objects and are widely employed in automated visual inspection solutions. With the advancement of CMOS technology, industrial CMOS cameras are increasingly popular due to their low cost and power consumption.
(2) Interfaces:
The front of an industrial camera is used to attach lenses and they usually have standardized professional interfaces. On the back, there are generally two interfaces: a power interface and a data interface.
Industrial camera interfaces include USB 2.0/3.0, CameraLink, Gige, 1394a/1394b, CoaXPress and others. Here, only a few common types are presented.
USB Interface:
Supports hot-plugging, ease of use, standardized and unified, connects multiple devices and can be powered via USB cable.
However, it lacks a standardized protocol and has a master-slave structure, with high CPU usage and unguaranteed bandwidth. USB 3.0 interfaces can be self-powered, but an external power supply can be used if USB power is unstable.
Gigabit Ethernet interface:
Developed based on the Gigabit Ethernet communication protocol, it is suitable for industrial imaging applications, transmitting uncompressed video signals over a network.
Offers good expandability, with data transmission lengths of up to 100 m (indefinitely extendable with repeaters), 1 Gbit bandwidth for instant data transmission, uses standard NIC cards (or pre-installed in PCs), is economical and uses inexpensive cables (standard CAT-6 Ethernet cables) with standard connectors. It is easy to integrate, cost-effective and widely applicable.
CameraLink Interface:
A serial communication protocol using LVDS interface standards, known for its high speed, strong anti-interference capabilities, and low power consumption. Developed from Channel Link technology, it adds some transmission control signals and defines related standards. The protocol uses MDR-26-pin connectors, offers high speed with bandwidth up to 6,400 Mbps, strong anti-interference capabilities, and low power consumption.
Gige interfaces simplify multiple camera setup, supporting 100 meter cable output. The Camera Link interface is designed specifically for high-speed image data needs. USB 3.0 interfaces are known for their simplicity and real-time capabilities.
Currently, the most used interface in artificial vision is the Gige (Ethernet) interface, which offers significant advantages over other interfaces in terms of transmission speed, distance and cost.
(3) Resolution
Resolution is a key factor in camera selection. It is important to understand the relationship between resolution, pixels, accuracy, pixel size, and sensor size as these terms are often confused.
Camera resolution refers to the number of pixels captured in each image, indicating the total number of light-sensitive chips, typically measured in millions and arranged in an array.
For example, a million-pixel camera might have a pixel matrix of WxH = 1000×1000. Pixel size varies between different devices, with each pixel having a specific position and assigned color value. The arrangement and color of these pixels determine the appearance of the image.
(4) Sensor Size
Sensor sizes (CCD/CMOS) can be confusing as terms like 1/1.8 inch or 2/3 inch do not refer to any specific dimension or diagonal size of the sensor, making it difficult to conceptualize its actual size.
| Sensor type | Diagonal line (mm) | Width (mm) | Height (mm) |
| 1/3” | 6,000 | 4,800 | 3,600 |
| 1/2.5 | 7,182 | 5,760 | 4,290 |
| 1/2” | 8,000 | 6,400 | 4,800 |
| 1.8” | 8,933 | 7,176 | 5,319 |
| 2/3” | 11,000 | 8,800 | 6,600 |
| 1" | 16,000 | 12,800 | 9,600 |
| 4/3” | 22,500 | 18,800 | 18,500 |
Target surface area = sensor width x sensor height
Sensor size affects field of view and working distance. With larger sensors and the same pixel density, the size of the pixels increases, improving the light-sensitive area of each pixel and improving image quality. Under the same working distance and lens, a larger sensor can capture a wider field of view.
(5) Pixel size
With the camera resolution and sensor size, the pixel size can be calculated:
Pixel size = Sensor size/Resolution (number of pixels)
This produces the pixel size in width and height.
Pixel size refers to the actual physical size of each pixel in the chip's pixel array, such as 3.75um x 3.75um. To some extent, pixel size reflects the chip's responsiveness to light. Larger pixels can receive more photons, producing more electrical charge under the same lighting conditions and exposure time.
This is particularly relevant for low-light imaging, where pixel size is an indicator of chip sensitivity. It's crucial to distinguish this from camera resolution: lower resolution values indicate higher resolution, while larger pixels imply higher sensitivity. These are two distinct concepts.
(6) Accuracy
Accuracy refers to the size of the actual object represented by a single pixel, expressed in (one*one)/pixel. It's important to note that pixel size is not the same as accuracy.
Pixel size is a fixed characteristic of the camera's mechanical construction, while accuracy is related to the camera's field of view and is variable. The lower the precision value, the higher the precision.
The size represented by a single pixel = Field of view width / Width resolution = Field of view height / Height resolution
Additional note: Considering distortion at the camera's edge of view and system stability requirements, we generally do not equate a single pixel unit with a measurement accuracy value.
Sometimes, depending on the light source, the calculation value is increased. With backlight, the accuracy is 1 to 3 pixels, while with a direct light source it is 3 to 5 pixels. For example, using a 500W pixel camera with a resolution of 2500 2000 and a field of view of 100mm 80mm:
- Single pixel size = 0.04mm
- Backlight accuracy = 0.04mm ~ 0.12mm
- Direct light accuracy = 0.12mm ~ 0.20mm
It is important to understand that when calculating resolution based on known accuracy, a camera with a higher resolution than the calculated value is often needed to meet the requirements.
(7) Image resolution
Image resolution is relatively simple to understand. Refers to the number of pixels used to display an image per unit of distance, similar in concept to precision but expressed differently.
Basic Selection Principles
When the field of view, that is, the size of the target, is fixed (the size of the target is generally considered as the field of view when selecting a camera), the higher the resolution of the camera, the higher the accuracy and resolution of the image .
When the field of view is not fixed, cameras with different resolutions can achieve the same accuracy. In these cases, choosing a camera with larger pixels can expand the field of view, reduce the number of photos needed, and increase testing speed.
For example, if one camera has 1 million pixels and another has 3 million pixels, and both have the same clarity (20um/pixel in precision), the FOV of the first camera is 20mm×20mm = 400 square mm, while the FOV of the second camera is 1200 square millimeters. If capturing the same number of targets on a production line, the first camera may need to capture 30 images, while the second camera only needs to capture 10.
