How to use machine vision for AI | TECH + Mynavi News Mynavi

Figure 1. Applications that require color imaging

Machine vision systems can use simple monochrome images to identify basic objects, but color images convey more information than monochrome images. Color becomes more important when AI is used to analyze the scene.

Adding color to machine vision is expected to bring a new dimension to automation in the new applications mentioned above. Colors make contrasts and differences stand out between objects in the scene. AI systems will be able to take advantage of these features to improve accuracy.

In this case, managing overall power capacity becomes even more important, especially for battery-powered, always-on devices. This could include devices that are expected to run on a single coin battery for more than 5 years. The most important design criteria for system power are image sensors, control systems, and communication interfaces.

Onsemi's RSL10 smart shot camera was developed to give engineers access to a complete low power image capture platform connected by BLE.

The latest version of this platform has added support for color image capture using a camera module based on the ARX3A0 CMOS image sensor. This module complies with the onsemi Image Access System (IAS) design format and allows module replacement using standardized connectors and layout configurations.

Recent RSL10 smart shot color cameras now support color-enabled CMOS image sensors. The platform has also been downsized and power is more optimized. After assessing the performance of the system, customers can use the design file containing the software to move on to developing their own smart machine vision IoT sensor.

Power management is important to maintain battery life. The RSL10 SIP (System in Package) / ARX3A0 uses a dedicated power management IC (PMIC) "FAN53880" for color cameras and a hardware-based smart power management mode. The power consumption when a color camera is connected by continuous image capture is 136.3mW. The power consumption drops to 88.77 μW when waiting for a trigger event, and the power consumption in power-down mode is only 30.36 μW.

As a result, if you capture one image per day, the color camera platform is expected to be able to operate for over 11 years on a single 2000mAh battery.

Event-triggered machine vision

The RSL10 smart shot camera is designed to provide event-triggered image capture. This means that the image is captured based on a predetermined event, rather than constantly streaming the image data. Event status is monitored using advanced sensors integrated into the camera platform.

Conditions that can be monitored using these sensors include motion, temperature, time, humidity, and acceleration. Developers can use the outputs of these sensors to create complex conditions for events. When these conditions are met, the RSL10 smart shot camera will trigger an image capture. The captured image is transferred to the smartphone or gateway via BLE.

The move to color image capture means an increase in the amount of data transferred over Bluetooth, but onsemi engineers are now able to handle it with little increase in system power requirements. Part of the key to this is the RSL10 SIP, a low power system in a package (SiP). This small, low-power SiP acts as a hub for the entire system, controlling the image sensor processor, driving the environment sensor, and managing BLE communication.

Cooperation with cloud AI platform

onsemi offers a custom mobile app for Android and iOS in addition to the RSL10 smart shot color camera. This will allow you to use Amazon Rekognition with your connected smartphone and a valid AWS account. Once your AWS account is connected to the RSL10 Smart Shot mobile app, you can upload and analyze images. When the analysis is complete, Amazon Recognition returns a percentage-accurate number of all the objects recognized in the image.

onsemi also worked with Avnet to integrate the RSL10 smart shot camera within the IoT Connect Platform, a cloud-based solution powered by Microsoft Azure. Designed to minimize the complexity of the IoT design process, IoT Connect provides a means of linking information from the camera platform to the cloud, enabling data interpretation, manipulation, and learning (AI). Customers can use this recipe to customize their own Proof of Value (POV) for their vision needs, such as object detection, analog meter readings, and inventory level checks. This will enable IoT projects to be validated more quickly and customers to enter the market faster.

Conclusion

Vision sensing is an exciting technology that is applied in many areas such as factory automation and agriculture. The benefits of having an "inspector" who never gets bored, tired, or making mistakes are important, but the mix of AI and machine learning makes a difference.

The RSL10 smart shot camera provides OEMs with a design platform that combines connectivity, color and monochrome images, and AI-based processing. With optimized low-power operation, these devices can not only provide access to advanced AI and machine learning through integrated cloud services, but can also operate for over a decade.