- This editorial is filed under:
- Automotive Component
- Consumer Goods/Appliances
- Life Sciences Pharma Biomed Medical Devices
- Rubber & Plastics
- Off-Road/Heavy Equipment
- Robot Manufacturing
- Building Products/Materials
- Fabricated Metals
- Printing & Publishing
- Arc Welding
- Wood Products
Understanding Real-Time Machine Vision
by Matt Edwards, Director of Professional Services
IntervalZero Posted 04/06/2018
As industrial processing power increases, real-time machine vision systems become more powerful and affordable. Machine vision technology has become a key element in the manufacturing sector. The technology utilizes industrial image processing and cameras mounted over production cells and lines to visually read, direct or inspect products in real-time without intervention from an operator. Machine vision systems can have many cameras that capture, interpret and signal information individually. The cameras are administrated by a control system set under pre-determined requirements and tolerance levels. In manufacturing environments, system reliability translates to increased product safety and production.
Real-time machine vision components
A machine vision system is made up of different components. It requires a host computer or development machine that is used to execute software development like prototyping vision strategies, sample image acquisition, and application benchmarking.
How does it work? After establishing your imaging strategy, you download the KINGSTAR machine vision system code to an embedded and ruggedized computer capable of running an RTOS. The machine vision code allows you to make modifications and integrate other functionalities like remote monitoring, motion control, and data acquisition. Next, you run the code in the PXI controller system that runs autonomously on the development machine. The PXI system incorporates a real-time operating system to ensure reliability.
Vision algorithms are not deterministic; image content determines the image processing speed so data-driven algorithms are unbounded. For instance, performing an image analysis on three particles will appear faster as opposed to an image with eight particles. Results computation time increases as the image variance increases as well as when performing time-variant analysis and processing functions.
To avoid performance gaps due to lack of determinism, it is advisable to prepare for the worst case scenario. So, analyze an image containing a high number of particles to determine the image processing speed and response time.
Purpose of an RTOS in machine vision systems
Images under high-speed industrial applications must be captured, processed and the outcome used to actuate a pass/fail mechanism within a predetermined period. To accomplish this, machine vision systems must employ real-time operating systems that allow designers to predict system jitter accurately. An RTOS also predicts the time taken to accept and process tasks like capturing images, processing them and I/O control. However, general-purpose operating systems such as Windows are also utilized to implement graphical interfaces.
Building your real-time machine vision system with KINGSTAR products ensures that you enjoy a completely deterministic control environment. This is because the system utilizes a simplified kernel real-time operating system to ensure fewer instances of latency or delays resulting from interrupted service routines. Additionally, KINGSTAR real-time machine vision systems integrate effortlessly with other automation and measurement devices. That means you can utilize high-speed interconnects between devices to enjoy advanced measurement capabilities as well as facilitate communication between operational devices.