How to choose these image sensors? Get these five key points first√

Infineon / Mitsubishi / Fuji / Semikron / Eupec / IXYS

How to choose these image sensors? Get these five key points first√

Posted Date: 2024-01-22

With the development of science and technology, machine vision plays an increasingly important role in intelligent transportation systems (ITS). Whether it is automatic license plate recognition, illegal parking detection, pedestrian detection, traffic flow analysis, or even driving violation identification, they are all inseparable from the support of high-quality image sensors. These sensors must have high resolution, high frame rate, wide dynamic range and low noise to adapt to various environmental conditions and meet the needs of different applications.

What are the characteristics of machine vision, and how does it compare to human vision?

While humans have always wanted to see full color and high resolution, this is not a requirement for a variety of machine vision applications. In machine vision, the image itself is not the most important. Most important is the information stored in the image: for example, the location and size of defects on a silicon wafer, the data stored in a barcode or QR code, or the numbers and letters on a license plate.

What are the rapidly growing application areas of machine vision?

ITS is one of them, especially in China. Cameras mounted on traffic light poles monitor license plate numbers and detect red light running and other violations. For speeding or breaking into specific areas illegally, fines are also digitally captured and even enforced. When a violation occurs, ITS records the driver's behavior, such as using a mobile phone while driving. Cameras for ITS applications typically require at least 1,000 to 1,500 pixel level resolution per channel, and some ITS vendors are moving to 2,000 pixels to prepare for the future. Requirements for vertical resolution are also growing, as ITS users want to know more about what the driver is doing, whether the pedestrian is in the crosswalk, and the status of nearby traffic lights.

Flat panel display inspection is another growing area, and inspection occurs at different levels, including checking whether each LED is turned on and off during the manufacturing process, and checking whether all LEDs are illuminated uniformly during final testing. There are other application areas including solar panel inspection and silicon wafer inspection. In addition, ON Semiconductor sensors can also be used for robotic operations in automated warehouses.

What improvements are required to specifications such as resolution, dynamic range, and frame rate?

These specifications apply primarily to moving targets. Camera manufacturers want higher resolution for applications such as silicon wafer inspection and solar panel inspection. The human eye has a high dynamic range, so one would naturally want to replicate this property in an image sensor. Currently, 12-bit resolution is sufficient for most high-performance machine vision applications, but some applications have begun to explore 14-bit solutions.

Also note that the combination of high resolution, high dynamic range, and fast frame rates generates large amounts of data. ON Semiconductor provides standard interfaces to transmit data from our sensors, and customers then need to provide the data processing capabilities to obtain meaningful results from the data. It's also worth noting that there's still a lot of demand for low-performance sensors. For example, barcode scanning does not require a 50-megapixel 12-bit color sensor.

In addition to resolution, dynamic range, and frame rate, what other specs should customers pay attention to?

Another key specification is power consumption. Keeping sensor power consumption low can extend battery life in portable devices, especially at high frame rates. In addition, power consumption means heat, and in image sensors, heat means noise, so keeping power consumption low is beneficial to improving overall image quality.

What other characteristics need to be considered when selecting a specific image sensor?

When designing a camera, choose a sensor family that can support multiple resolutions and configurations as well as IP reuse. For your application, a telecentric configuration may be sufficient. If a telecentric configuration is not sufficient, look for sensor families with chief ray angles (CRA) optimized for specific use cases (such as ITS).

An example of this family is the ON Semiconductor XGS image sensor. The XGS sensor is available in 11 resolutions, ranging from 2Mp to 45Mp, with one size supporting resolutions from 2Mp to 16Mp and another from 20Mp to 45Mp. All versions offer true global shutter and 12-bit analog-to-digital converters. The version for the ITS market provides optimized CRA adjustments. All XGS sensors feature low power consumption, in some cases dissipating half the power of competing devices.

What will image sensors look like in the near future?

For high-end machine vision applications, the pursuit of higher resolution will still be the goal. Optical formats will become larger and pixels will need to become smaller to keep sensor size manageable and allow for the use of lower cost lens systems. Large sensor size requires higher data readout performance. As the cable length doubles, the connection line parasitics will quadruple. On the manufacturing side, higher-density sensors are inherently more susceptible to artifacts, and yield optimization is critical to reducing costs.

In terms of applications and markets, panels and displays are everywhere from mobile phones to car dashboards to 80-inch TVs, driving demand for sensors and systems to monitor them and leaving huge room for development.

#choose #image #sensors #key #points