Trends In Machine Vision Camera Technology

Cost-Effective SWIR Imaging

Rapid advancements in sensor and automation technology are bringing significant changes to the machine vision industry, optimizing industries such as manufacturing, healthcare, autonomous vehicles and robotics through artificial intelligence (AI), vision software, and novel hardware architecture. These changes increase productivity, cost savings, and improve decision-making capabilities in a number of applications across all industry spaces.

Edge Artificial Intelligence for Machine Vision

Edge AI, or edge learning, is a particular type of artificial intelligence that uses algorithms and deep learning neural networks directly with computing devices at the edge of a network. Data processing occurs primarily onboard the camera and information is exported directly to the connected cloud. These systems have the benefit of automating repetitive processes and operations for optimized power consumption, network latency, and increased overall application efficiency.

A LUCID Vision Labs™ SENSAiZ Intelligent Vision CMOS Camera keeps track of inventory levels at a convenience store and alerts personnel when supplies drop below a specified amount.

High-Resolution Artificial Intelligence for Machine Vision

Artificial intelligence in and of itself is not a brand-new trend. Deep learning neural network models are being applied to applications at an increasing pace to achieve incredible results. What makes the LUCID Vision Labs™ SENSAiZ Intelligent Vision CMOS Camera different is the unique combination of high resolution with AI. Previous iterations of AI for machine vision utilized slightly lower resolution sensors, sampling the image at VGA resolutions.

Event-Based Machine Vision

Event-based machine vision, also known as neuromorphic vision, is an imaging method where the camera sensor, sometimes referred to as a dynamic vision sensor (DVS) takes a continuous record of exposure intensity, unobstructed by shutter. On a per pixel level, changes in intensity are recorded asynchronously and in a parallel manner similar to neural networks, conserving limited on-board computational resources.

Advancements in SWIR Imaging Bring Affordability

Short-wave infrared (SWIR) imaging uses light typically in the waveband between 0.9 and 1.7µm but can also include light from 0.7 to 2.5µm. Because SWIR wavelengths are outside of the visible spectrum and typical silicon sensors used for visible light are only sensitive to light up to around the near infrared spectrum (between 650nm and 1µm), SWIR sensors are constructed with other materials including indium gallium arsenide (InGaAs) and indium phosphide (InP). Traditionally, these sensors were very difficult to manufacture yielding high price points. However, advancements in SWIR sensor manufacturing techniques in the past few years have significantly improved manufacturing efficiency, and thus their affordability.

The quantum efficiency (QE) of traditional silicon (left) sensors is only sensitive to around 900nm to 1µm but InGaAs are sensitive out to much farther as seen in this visual-SWIR InGaAs hybrid sensor (right).

Article Source:  https://www.edmundoptics.com.tw/knowledge-center/trending-in-optics/trends-in-machine-vision-camera-technology/