Network technology of the future for more security

Deep Learning Processing Unit (DLPU) and intelligent algorithms for more health protection and to increase occupational safety

Project planning

On drilling platforms, employees are at risk if they are in a certain danger zone around the drilling device while the drilling device is rotating. To increase employee safety, so-called red zone supervision, i.e. monitoring a danger area, makes sense. The so-called people counter can count whether there are still people in the danger area and, if this is the case, prevent the drill head from being switched on. Dangers can be drawn to attention through an acoustic alarm and/or signal lights.

Cameras have long been used for more than just imaging surveillance. Due to the ever-improving processors and the associated performance of the cameras, the decentralized task of image evaluation can now also be carried out directly on the camera - so-called on edge solutions. Today's network camera technology is already so advanced that images that are displayed are preselected before they are displayed. There are questions in the room such as: Is something moving? Is this an interesting process? Is someone stepping over a fence? Does someone step over a line? Something like this is already solved algorithmically today.

Motion detection (video motion detection) has long made it possible to detect people or vehicles in certain areas. However, the algorithmic analysis option reaches its limits because it is not capable of learning. AI systems close the borders, AI systems are adaptive algorithms - also in camera technology. Such technologies also take our camera technology to a new level. The question is no longer “Can the camera see this?”, but “Can the camera learn to see this?” This is exactly what our customer wants. He wants the camera to be able to detect whether someone is wearing a safety vest; whether someone is wearing a helmet; whether a person is lying on the floor; are everyone there? Did a person crash? Is a person injured? With the new technology (intelligent algorithms and deep learning processor unit) it is possible to distinguish between a person or a car in the image. Object analytics is interesting, for example, where the entry of a person into a danger area in which a machine is moving needs to be detected and reported. Using suitable software, it would be possible to determine whether a person is standing too close to the edge or whether someone is wearing a helmet. Our intelligent cameras have the necessary hardware and programs from third-party users can be easily integrated.

The camera system should make experiences and learn from them. Significant technological advances have been made in recent years, but we are still at the very beginning. As is the case with new technologies: at the beginning you rely on infrastructure. Our intelligent cameras are equipped with double CPU. While the main CPU deals with video streaming, the Deep Learning Processing Unit (DLPU) is exclusively intended to execute intelligent analysis algorithms. We have conceptually developed a possible scenario for helmet detection and vest detection, and have also upgraded our technology to the extent that later AI algorithms can be installed on the CPU. Together we take a visionary look into the future and ask ourselves how AI-supported video surveillance systems will make work safer in the future.

Our sequence chain has become “See - Analyze - Act - Learn - See - Analyze – etc...” It no longer ends with the action, but with “learning and experiencing” it is close to the processes that take place in our brains.
Safety, performance, accessibility and sustainability are fundamental concepts that must be implemented.

• Safety – Is there a movement in the dangerous zone? Warning and locking, protecting your own safety, your employees and the entire company
• Performance – progressive scanning, high quality image for identification
• Accessibility – Accessible by multiple authorized users, at any time, from any networked location worldwide
• Sustainability – Configurable, self-optimizing

https://www.samcon.eu/en/company/report-ai/

Testimonial

“The ability to integrate AI-supported, safety-related software into our camera system was very important to us. SAMCON has joined in and developed and certified the first explosion-proof deep learning camera especially for us.”