Latest technologies
DEEP LEARNING – KEY TECHNOLOGY FOR AI
Just imagine that objects and people in the warehouse are being automatically recognized with
their
essential characteristics by a camera in real time. A wide range of innovative application fields
is
available: starting from automatic recording of material flow, stock control, maintenance of master data
together with support for the staff during entry and quality assurance, privacy protection in video
material, access control and time recording up to camera-guided AGV’s. These benefits lead to maximized
quality and reduced costs.
In logistics a lot of time is spent on identification (scanning) of products, packages, pallets etc. and
on checking quantities or exchanging falsely picked products due to high error rates caused by manual
identification.
Logivations set its goal to minimize the time spent on object identification and to reduce error rates to
achieve a significant efficiency increase.
Methodology and process
Cameras & algorithms instead of Scanners and RFID for object identification
Why haven’t others done this yet?
Making people unrecognizable in videos is hard. Once a person’s face is visible - even for less than a
second – this person can be easily identified by a human watcher. Thus, a computer algorithm needs to
look for face templates and try to recognize these templates in the pictures. Significant complexity is
not only about the object itself, but also about its position, location and lighting conditions
Conventional solutions are based on a few predefined templates. They are not comprehensive in considering
these situational factors. That is why they can´t offer the necessary reliability to control such
crucial cases as e.g. privacy protection, access control or automatic execution of the processes (see
example below):
The Deep Machine Learning algorithm from Logivations practices like a human being and further learns
after each positive recognition, instead of just being limited to predefined templates. As a matter of
fact, it is trained to recognize objects and faces based on real examples.
We succeeded in providing exactly this reliable recognition in realtime and out of the Cloud by using
Machine Learning and neural networks from the current developments. With this our object and face
recognition goes far beyond the usual functional spectrum of similar solutions, as besides detection it
can also capture quantity, positions and characteristics of objects.
Let’s get technical
Here’s how most current solutions work: They use the so-called Viola-Jones algorithm. Typically on a
face, the eyes are darker than the nose, thus the algorithm looks for such patterns:
Once a part of an image contains such a pattern, it is classified as a face. This works well for faces
with no rotation, and as a bonus is quite fast.
But, what happens with faces that can be viewed from the side? Here, only one eye is visible, so the
pattern won’t be matched, and no face is detected. Thus, another pattern has to be added for profile
faces. This is already quite hard, as there is no simple pattern applicable to every profile faces.
Nearly impossible however is finding patterns for unusual head positions: someone that is looking at the
ground, or up a wall. These situations are very hard to recognize with the Viola-Jones algorithm.
That’s the issue: We, as humans, can recognize a known person even if we can only see a profile view of
their face. Thus, an algorithm useful for privacy protection has to find all those cases as well and
blur them.
Game-changers: neural networks
Logivations set out to build a better solution. With latest technologies, such as neural networks we used
a database of more than 500.000 images of humans to train an algorithm to find patterns for heads on its
own – in all positions, rotations and lighting conditions imaginable. Here’s an excerpt of the patterns
found:
As one can see, instead of black-and-white pixels, our algorithm found lots of complex patterns. Thus, it
can detect heads even in very non-usual positions. This solution easily detects heads where other
algorithms fail
In these images, 0.997 means the algorithm has a 99.7% confidence in the detection.
Such an algorithm requires considerable amounts of resources. Using the power of the cloud we can provide
a fast and seamless processing of the videos.
Technology and background
- Intelligent Machine Learning algorithms for object recognition in pictures/videos (“neural
networks”)
- Significantly lower hardware costs of high-resolution cameras, tablets/smartphones
- Enormous increase in computing power during recent years (GPU computing)
- Integration into warehouse management systems, standard Interface to SAP available
- Algorithm is trained on sample images then automatically extracts (“learns”) required patterns
- Logivations Deep Machine Learning can be used for fast and robust detection with standard cameras
VISUALIZATION IN 2D, 3D & VR
Visualization of the complete warehouse in 2D, 3D and even in VR enables to have deeper insights compared
to using only data and presentations since it gives impressive inter-divisional views in processes,
optimization ideas and their results. Labor efforts, load on conveyors or even an ABC distribution of
the stored goods deliver powerful visual support for important decisions. They also improve workshops or
meetings and save time at the manual preparation. Furthermore, big data can only be managed properly
with help of visualization.
- Balanced utilization of staff and all resources
- Prevention of peaks
- Holistic reduction of labor efforts
- Exact forecast of needed staff in the future
- Delivery in time, even with high order dynamic