Industrial automation trends and the rise of artificial intelligence (AI) are driving the rapid growth of the machine vision market. Edge AI continues to spawn new frontiers. The new crown epidemic has further accelerated these automation and intelligent trends. The traditional machine vision industry is also quietly changing, and these are inseparable from the eyes of image sensors. ON Semiconductor is a semiconductor leader in intelligent perception, especially in the world of industrial machine vision, and is also far ahead in the edge AI market. It has a broad lineup of imaging solutions and advanced imaging technologies to meet different application needs.
How Industrial Imaging Applications Enable Fast, Accurate Decision Making
There are common industrial application needs such as measuring the size of an object or determining the level of a liquid, providing a robot with “eyes” when positioning or guiding to determine where to grab an object, counting or inspecting a quantity or judging whether a package is complete , to be able to read barcodes or perform optical character recognition when decoding or identifying, and so on. Designers need to consider and optimize different imaging parameters based on application needs: Resolution is directly related to the number of pixels available on the image sensor, distinguishing individual measurand features in the scene. High image quality is essential to eliminate false or inaccurate results. A high frame rate is good for capturing high-speed moving objects and trajectories, and can solve problems in a timely manner.
In addition, the cost of the entire system is a key factor in bringing all these parameters together and must also be considered. Increasing resolution simply by adding more pixels results in a larger image sensor that costs more, but also requires more expensive lenses. If the resolution is increased by reducing the pixel size, the increase in sensor and lens cost may be reduced, but this may also affect image quality. If the resolution and frame rate are higher than what a given application requires, then the application’s data bandwidth requirements (and infrastructure costs) will be higher than otherwise.
Therefore, the key is to intelligently improve overall imaging performance to continue to meet the growing demands of these applications.
Imaging technology: high speed, short exposure
The visual inspection of the factory is basically corresponding to high-speed moving objects, and the speed of the conveyor belt is particularly fast. Therefore, high frame rate and short exposure control are used to prevent motion blur, and AI algorithms can be used to achieve correct intelligent judgment and rapid decision-making.
Imaging Technology: Global Shutter
Because there is no storage node that can store charges inside the rolling shutter, the data of this line needs to be read out as soon as possible at the end of each line of exposure, and then the exposure of the next line starts. Then, when the object is moving rapidly, the whole frame of the image has different time from top to bottom. The exposure method will cause spatial distortion and deformation of the image, while the global shutter exposes the entire frame at the same time, and all the captured pixels correctly restore the true appearance of the moving object, so there will be no distortion.
A global shutter image sensor captures images and stores the captured data on storage nodes on a pixel-by-pixel basis. In very bright light conditions, it is critical to protect storage nodes from ambient light corrupting data while keeping data on storage nodes. One way to measure the protection of this node is called Global Shutter Efficiency (GSE). Poor GSE can lead to ghosting effects in the image. This ghosting effect can cause performance issues for machine vision algorithms. ON Semiconductor’s global shutter image sensors offer the industry’s best GSE.
Imaging Technology: Resolution
Flat panel inspection is the most challenging application for image sensors in the entire industrial machine vision industry. It goes from 1K, 2K, 4K all the way to 8K, and the pixels are gradually expanding. There are roughly two types of detection: one is dark detection, which mainly detects some fingerprints, scratches and other physical problems before power-on; the other is after power-on, especially in OLED, plus the latest AMOLED (AcTIve Matrix OLED) . LCD has a bright panel behind as the light source, while OLED, especially AMOLED, each pixel of it is a separate light source. The luminous intensity and color uniformity between pixels must be accurately detected, which requires very high image sensors. In the past, detecting 1 pixel on an LCD panel requires 9 pixels (3×3), while OLED may require 16 pixels (4×4) or even 25 (5×5) pixels. So the demands on image sensor resolution are getting higher and higher, and inspection cameras are required to provide very high image quality and high uniformity – to ensure that quality issues in the camera are not misinterpreted as product defects in the display.
Inspection of printed circuit boards also often requires high-resolution image sensors to confirm that circuit board components (ICs, capacitors, resistors, etc.) have been properly placed and soldered in place. The speed of inspecting a board is a combination of resolution and frame rate—the larger the size of the image capture, the more boards can be inspected at one time; the faster the image capture, the more efficient the inspection.
Using ON Semiconductor’s XGS 45000 captures a much larger image: (over 8000 pixels wide) at a data rate of almost 1900 pixels per second, three times more efficient than previous solutions. This combination of high resolution and high speed enables this type of inspection application to be performed faster and more efficiently, increasing productivity in the manufacturing process.
Another example is broadcast/surveillance applications, which combine many different performance vectors, require high resolution, high bandwidth, and high image quality, and the demand for resolution continues to grow, from HD to 4k to 8k. ON Semiconductor’s XGS 45000, which actually has a higher resolution than 8K, can provide 12-bit data output at 60 fps, and capture slight movements clearly. It is worth mentioning that the XGS 45000 camera will be used in the Tokyo Olympic venue.
Imaging Technology: Image Scale
Not all applications require high resolution, and some machine vision applications also have specific resolution requirements. Unlike traditional RGB viewing applications driven by display standards (16:9 or 4:3), many robotics and machine vision scenarios can be optimized by using different resolutions. For example, ON Semiconductor’s 2-megapixel AR0234 gives extra pixels in the X direction for a better barcode scheme. In the XGS series image sensors, 8 million pixels, 9 million pixels, and 12 million pixels provide image ratios of 2:1, 1:1, and 4:3, respectively, and XGS’ 30 million pixels and 320 million pixels provide 1: 1 and 4:3 image ratio, 1:1 can be used for semiconductor inspection, and 4:3 can be used for screen inspection.
New Imaging Requirements: Stereoscopic Vision
In the past two years, many applications have increased the demand for depth information. In the past, 2D imaging information began to be promoted to 3D stereo vision. Binocular cameras and cameras combined with structured light are relatively common 3D application solutions. ON Semiconductor’s PYTHON series of image sensor products are often used in these solutions. There are also 3D solutions using time of flight (TOF), Super Depth pixel and radar. ON Semiconductor’s direct time-of-flight (DTOF) lidar products include single-point silicon photomultipliers (SIPMs) for medical, industrial, and automotive applications, as well as 1×12 or 1×16 arrays, next-generation Lidar will use such products. The newly developed 400×100 single-photon avalanche diode (SPAD) area array is not just a point cloud, it is almost the same as an image sensor, and can realize images with depth information.
The following introduces several imaging solutions of ON Semiconductor:
Industrial global shutter solution: XGS series
XGS is ON Semiconductor’s latest generation of industrial-grade global shutter image sensor, using the latest 3.2 um pixel process, with good image quality, low noise, good consistency, high frame rate, low power consumption, wide operating temperature, and extremely Cost-effective advantage, can be used in a variety of industrial applications. It is a serialized family product X-CLASS platform, from 2 million pixels to 45 million pixels, it can be compatible with a total of 11 resolutions with two sets of hardware circuit boards, of which 2 million pixels-16 million pixels can be easily put into 29 The camera platform of mm*29 mm is pin-to-pin compatible with 20-45 million pixels. Customers can design a hardware platform and select the corresponding resolution sensor according to the specific application requirements. Camera manufacturers can leverage existing parts inventory and accelerate time-to-market for new camera designs. This generation of products will subvert the current market competition and provide users with a very attractive choice. Specifically, the XGS 45000 is used for high-resolution machine vision and industrial imaging applications that require high-speed data capture and transmission, and the XGS 12000 is a cost-effective choice, capturing clear, high-resolution images at 90 fps.
Economical global shutter solution: AR0234
AR0234’s 2 million pixel resolution, 120 frame rate and optical size are suitable for most traditional industrial inspection, autonomous mobile robots, industrial code scanning, binocular depth measurement, etc. ON Semiconductor also uses AR0234 with system-on-chip (SOC) image signal processor AP1302 to make corresponding solutions to help customers speed up development.
Facial Recognition Payment Application: AR0230 HDR
Contactless passage through gates such as airports, railway stations, and high-speed rail stations will become more and more common. AR0230 is the only 2-megapixel image sensor in the industry with on-chip bandwidth dynamic synthesis function. Wide dynamic range is very important for contactless systems. The AR0230, on the other hand, can provide very good image quality and support various lighting conditions from dark nights to bright sunlight, thereby improving the judgment accuracy of artificial intelligence.
Back-illuminated rolling shutter: AR0521
The AR0521 can capture 5-megapixel images at 60 fps, and ON Semiconductor also offers a NIR-enhanced version. The 5-megapixel resolution is suitable for various applications such as security, industrial, surveillance, and edge AI. For example, the reference design of thermal energy meter using AR0521 exhibited by ON Semiconductor at Vision China 2021 is used for non-contact identification of whether people passing by in front of the camera have fever and whether they are wearing masks, which is helpful for investigating epidemic cases and reducing the spread of the virus.
A Platform That Simplifies System Design: DevSuite
For easy customer evaluation, ON Semiconductor provides a PC-based evaluation system, DevSuite. Designers can load DevWare software onto a personal computer and connect to the sensor hardware of choice through a standard USB interface. With this platform, only the hardware of the sensor board needs to be replaced to evaluate the performance of the sensor.
Application Cases of Image Sensors in Industry
1. XGS series for intelligent transportation system, machine vision and industry
JAI A/S, one of the leading suppliers of industrial production inspection camera solutions, has integrated the XGS 45000 into its new 45-megapixel industrial camera. The SP-45000M-CXP4 offers 44.7-megapixel black-and-white resolution at 52 frames per second, leading the market, and supports full 8K resolution at over 60 frames per second.
Basler, the world’s leading manufacturer of digital cameras for industrial and retail applications, medical equipment and transportation systems, has also integrated XGS sensors into their boost platform based on the CoaXPress 2.0 machine vision standard, saying the XGS series is ideal for new designs and The transition from CCD to CMOS in existing high-resolution vision systems.
Teledyne Imaging, one of the world leaders in digital imaging technology, also values the performance, higher resolution and quality of XGS sensor technology, using the XGS 45000, XGS 30000 and XGS 20000 in its GenieTM Nano-5G M/C8100 area scan camera.
2. The Perseverance rover lands on Mars
Perseverance is the first mission to actively navigate during descent, a huge advance in Mars mission technology, powered by image sensor-based cameras. Perseverance has a total of 23 cameras, 6 of which use ON Semiconductor’s image sensors, including the 1.3-megapixel, 1/2-inch CMOS image sensor Python 1300 and the 5-megapixel, 1-inch image sensor Python 5000.
Summarize
Future image sensors will still pursue high resolution, high image quality, and high cost performance. As a leader in machine vision, ON Semiconductor will continue to innovate, with 3D imaging, hyperspectral and multispectral imaging as the future direction, empowering machines to surpass human vision.