The software behind our breakthrough visual guidance system.
A category-shifter, it redefines the automated process by giving blind robots true sight and hand-eye coordination. Paired with our hardware systems, Robeye and RAIO, it quickly upgrades any automated line.
Developed to mimic the human visual process, our software uses a unique algorithm that gives production robots the ability to view, recognize and locate pre-determined parts. In any situation where an object needs to be grabbed using a robotic arm, our software can be taught to see it, recognize it, then grab it every time—no matter the object’s position or orientation.
With CortexRecognition, we are the first to solve some of robotics’ longstanding challenges, including the ability to recognize thousands of unique items and perform random part picking and sorting.
An Industry First: 6D Technology with a 2D Camera and Zero Engineering
CortexRecognition calculates six degrees of freedom with one image—and no three-dimensional camera technology. Eliminate lengthy set-up time and complex digital images. We are the first in the industry to use one eye, one image and no structured light to calculate depth, rotation and location with high accuracy and speed.
6D – Capable of up to six degree of freedom (X Y Z Rx Ry Rz).
Random Part Pick Multiview
Identifying and precisely picking random parts in a pile has been one of the greatest challenges of modern robotic guidance, until now. Our innovative technology allows your robotic workforce to identify unique and similar parts and pick them accurately every time, utilizing our breakthrough Multiview feature.
- Finds Objects in XYZ RxRyRz
- One Camera No Calibration
- No Structured Light
- Automatic Robot and Camera frame alignment
- Sub-Millimeter Accuracy in XYZRxRyRz
- Works with complex 3D Parts
- Works with every Robot (TCP or Serial)
- Works when competitors are failing
Easy and Innovative Integration:
One Camera. One Cable and Zero Engineering.
Our system upgrades nearly any automated workforce within hours and keeps running for decades with zero engineering.
Any Lens. Any Standoff. No Calibration.
What if the object is too small? Too large? Too close or too far away? What if you needed a different lens or a different standoff? At any length and with any lens, CortexRecognition will accommodate without calibration.
Our software has been developed with the ability to seamlessly communicate with almost any industrial robotic system, so integration with your production process is nearly effortless.
Show it the item to learn and within minutes CortexRecognition will remember and recognize the object, no matter its orientation. It’s that simple.
How to Guide:
- Position robot
- View object
- Push “Teach” button
- Erase shadows
- Measure distance to object
- Guide robot
1. Look at object
2. Press Teach
3. Erase Outliers
4. Ready to Guide
* this list is incomplete and growing
- Automotive: Picking parts from structured and semi-structured racks
- Automotive: Picking parts from a pile
- Automotive: Picking from AGV or conveyor
- Automotive: Picking deformable objects (chip bags like)
- Automotive: Geometric Best Fit and GEO-Assembly
- Automotive: Wheel Alignment
- Automotive: WheelHouse Hamming
- Automotive: Tire Deflation (looking for nozzle)
- Automotive: Piston Stuffing into engine block
- Automotive: Seal application
- Casting Industry: Picking up hot mold from furnace
- Casting Industry: Lancing
- At-A-Glance Applications
- Part-in-Rack Unloading
- Conveyor Picking
- Engine and Transmission Picking and Placement
- Gripping, Applying Seal, Assembly, Etc.
- Tire Valve Stem Inflation and Deflation
- Engine & Transmission Pick & Place
- Piston Stuffing
Frequently Asked Questions
How quickly can CortexRecognition identify and localize an object?
This depends on the size and level of detail which the object possess. Generally, automotive applications see acquisition times of less than 500ms for recognition, localization and data availability to the robot.
What object sizes can CortexRecognition work with?
Some objects that CortexRecognition works with are as small as 1 x 3 inches, and others are as large as the side body panel of an SUV!
What degree of accuracy can CortexRecognition achieve?
Our applications employ our patented iterative process wherein we ensure that the pose offset data that CortexRecognition gathers is correct. This iterative process ensures accuracy that is only limited by the capability of the robot with which CortexRecognition system is integrated. Currently, there is a production scenario that CortexRecognition is used in, where a robot is performing piston stuffing with an accuracy of 100 microns.
Does CortexRecognition need any additional pieces of software to function, and can it read output in real time?
CortexRecognition requires only a small host of third party drivers to function fully. Real time output reading has not yet been achieved due to latency between the robot and the CortexRecognition software suite.
Should certain objects be avoided by CortexRecognition (i.e: polished metal, glass, etc.)?
CortexRecognition system achieves the highest level of robustness when recognizing objects with many details. CortexRecognition system is least robust (in 6 degrees of freedom) when recognizing small, simple shapes that lack detail and contrast.
Does CortexRecognition require a CAD model of the objects that will be recognized and localized?
Absolutely no CAD is required. CortexRecognition uses simple point-and-click object teaching.