AI camera-based identification solution
represents the cooperation between ceiling recognition units (cameras with GPU) and digital twin for full area coverage, real-time transparency of all processes and easy reorganizing of structure, restrictions, or routes in a few clicks
Real time location is a key requirement for all kind of industry 4.0 applications. So far real time
location systems were based on UWB or RFID or similar technologies.
But there is an easier, much more powerful way! Real time location can be done like humans do: just
look around, see what is there, where it is, how it looks like and what it does. Object recognition
with Artificial Intelligence allows this straightforward approach now to be done by computers with
special, but standard processing units, so called GPUs. Combine this idea with a Digital Twin which
represents reality in a 3D computer model. This combination allows an extremely powerful location,
thus remaining simple and fast. AI can recognize a forklift and where on the picture the forklift is
located. So you can draw a straight line from the camera to the forklift. As we all know: forklifts
are not flying. They stand or drive on the floor. The Digital Twin can tell where is the floor in this
building – at the point of intersection of the floor and the straight line from the camera must be the
forklift. As AI can recognize very good the brand and model of the forklift the dimensions of the forklift
are known, this means the location can be very precise. And just one camera with a GPU can do the location
in cooperation with the digital twin. No triangulation needed!
If objects may look identical they get an AprilTag attached. This is like a number plate of a car and can
be printed with any printer.
Background for a Technology Revolution
Neural networks automatically "learn" the patterns required for object identification.
In practice, often only a few sample images are sufficient
Extensive databases with training data enable previously unreached identification rates
(outperforming human decision makers)
Once learned, the trained networks can be used for many tasks
Very fast processing: less than 0.1 seconds per image
Extremely fast development of this technology due to fast enhancement of the underlying hardware
Object localization with cameras, AI and Digital Twin
Real time location, tracking and mirroring to the Digital Twin
Supervision of storage areas, lanes or buffers with cameras and AI
Recognition of state of transporter: loaded – onloaded, no additional sensor needed
Both transporters or assets or goods can be located and tracked
Automatic booking in WMS when e.g. pallets are picked or dropped
Can be set up to be 100% privacy data protection compliant by blurring the faces of persons
High accuracy is achieved by automatic camera calibration – from 1 mm to 10 cm
Time-consuming scanning processes and sources of errors are eliminated
Optimal routing of the transporters from start to end – full navigation system
Highly distributed, decentralized and parallelized software architecture
Possibility of integration of all location data to ERP, MES or WMS systems
All functions can be selected easily using intuitive interface
Seamless switching between indoor and outdoor spaces
How does AI-Based RTLS work
The operations area is stored as a true-to-scale digital twin in W2MO
2D and 3D representation in digital twin to create real-time transparency of all processes
AI camera-based identification solution consists of two key components: Recognition units
(cameras +GPU) mounted in such way that the entire operations area is supervised and a digital
twin allowing to create and change structure, restrictions, and routes with a few clicks and
save all tracking data.
Fast camera installation based on digital twin camera plan to ensure full coverage of area
Low cost of recognition units
Recognition is decentralized, directly done on the recognition units – no streaming of videos,
only transfer of position data and properties of objects.
One camera per 100 m2 to 1.500 m2 – precision from 1mm to 10cm
Usage of AprilTags (like number plates of cars) to locate and identify similar looking objects.
AprilTags can be printed on paper, plastic or metal – low cost of the whole solution
Recognition unit with two cameras, metal housing and cooling element, robust, dust resistant and easy to mount.
Power supply and network connection by PoE
Advantages of an AI-based RTLS
Recognizes and locates any trained object, finds its properties and what the object is doing
(unlike UWB, FRID, Laser which determine only a tag)
Digital twin maps reality with digital model and allows many more applications e.g. fleet
management, staff planning, action triggering, security, collision prevention, social distancing
and much more.
Area supervision included, fill grades of lanes or buffers can be recognized, too
Digital Twin allows to “understand” the environment, e. g. if a barrier is between a forklift
and a person or the person is in danger
Persons can be either blurred in order to protect privacy or they can be located and tracked –
in order to measure productivity (e. g. for using a performance-based remuneration of workers)
Extremely high reliability and stability
Very affordable investment and long-term total cost of ownership
W2MO Machine learning can check if you’re social distancing properly at work during Corona-times
Monitoring the minimum distance – e. g. six feet – between people in a production site or in a warehouse
If the minimum distance is violated, a real time notification will happen, e.g., siren, headlights, flashlight
controlled via IP switch actuators, as well as mail notifications or smartphone warning tones (only if privacy protection is not used)
If desired: documentation for health authorities can be provided
Option to include or exclude privacy protection (according to GDPR)
W2MO tracks the movements of people and other objects and can calculate the distances between them. As soon as
people get too close and violate the minimum social distancing, a warning can be issued: optically, with lamps /
headlights, or acoustically. If desired, employees wearing / not wearing a mask can also be identified. This makes
it easy to monitor that only employees with a mask are working in certain areas.
Privacy protection can be handled differently depending on the customer's wishes. If the protection of privacy is
of great importance, people are automatically blurred on the detection unit. In this case, the system only warns
that two people have come too close.
To enable traceability in Corona times, individual people can also be explicitly identified and tracked. If there
is too close contact, the system can document who has been in contact with whom and for how long, thus effectively
supporting the follow-up of health authorities.
Detection takes place directly on the detection units (cameras with ML-stick) - no streaming of huge videos required.
The customer can install the detection units by themselves - the software can be provided remotely by Logivations.
A virtual project without physical contact is easily possible.
Automatic recording of goods movements
Automatic recording of storage and retrieval
For high-bay storage and floor spaces
Customer specific training of the load appearance possible
Can be combined with Logivations identification gate or onboard forklift cameras
No modifications or additions to the forklift required
In addition to tracking the forklifts, goods movements can also be tracked and automatically posted to the WMS. The storage
and retrieval of pallets can be recognized with the help of detection units. The current position of the forklift is known
from the recognition functions.
This means that all bookings can be made automatically when moving goods. Storage and retrieval can also be documented in
order to have seamless traceability, which means documentation of all goods movements is saved in case of investigations.
No modifications to the forklifts are required for the automatic booking of goods movements. No matter which manufacturer
and which year of manufacture - the system can always be trained so that goods movements are recognized.