Enhanced driving safety with AI camera for person detection
On busy construction sites, a construction machine equipped with the intelligent 2D/3D sensor system O3M AI can reverse more quickly and municipal vehicles can stop reliably, even if the person crossing is difficult to see. At airports or container terminals, the sensor system guides the centimeter-precise positioning of a passenger bridge or container without causing collisions. Annoying false alarms, which are almost unavoidable with conventional 3D systems, are virtually eliminated by combining the two technologies. The number of false alarms is significantly reduced. This is because the system recognizes whether a person or a pile of earth is in the path of travel. This increases the efficiency of the work process.
This is made possible by the unique combination of a 3D PMD sensor and a 2D AI camera in an embedded system for area monitoring. By fusing sensor data from two different sensor types on a fully integrated and pre-calibrated platform, the system gains significantly in reliability. AI evaluation of the 2D camera data improves the quality of object recognition and enables differentiation between people and other obstacles. The new collision warning system thus works much better and more accurately than conventional solutions. The O3M AI smart sensor system generates a video from the data of 2D and 3D cameras that reproduces the environment and simultaneously displays information on the exact distances of all objects and people within a radius of up to 25 m in the image. O3M AI can differentiate at any time whether emergency braking, controlled stopping, or slowing down would be appropriate in the respective situation.
The digital, high-resolution Ethernet camera is equipped with a powerful processor with AI accelerator (NPU). This enables ultra-fast processing of AI algorithms for person recognition directly in the system. In addition to the Ethernet camera, a 3D PMD camera is integrated. Based on the PMD time-of-flight principle, this provides an exact distance measurement to the object in each pixel. The depth image from the 3D system and the video from the Ethernet camera can be combined or used individually, depending on the application.
The O3M AI's person and obstacle detection has been optimized for mobile work machines. The collision warning system can accurately distinguish a distant obstacle from a person in the immediate vicinity, even if that person is lying on the ground, wearing dark clothing, or is difficult for conventional systems to detect due to large tools or unusual body postures; and it can do so in bright sunlight as well as at dusk and up to a maximum range of 25 m with an accuracy of 10 cm.
The AI, hardware, and software were developed in Germany by ifm. The examples used to train the AI are taken from the typical working environment of mobile machines and include a wide variety of person types, equipment, and working positions. This results in a much greater gradation of possible alarms and warnings issued by the system. This is an important criterion when it comes to striking a balance between efficiency and safety.
The new combination sensor system for mobile machines is available in different versions. With integrated AI and person recognition, ifm markets it under the item number O3M372. It is part of a comprehensive ecosystem for mobile machines that ifm has built up over 35 years.
On busy construction sites, a construction machine equipped with the intelligent 2D/3D sensor system O3M AI can reverse more quickly and municipal vehicles can stop reliably, even if the person crossing is difficult to see. At airports or container terminals, the sensor system guides the centimeter-precise positioning of a passenger bridge or container without causing collisions. Annoying false alarms, which are almost unavoidable with conventional 3D systems, are virtually eliminated by combining the two technologies. The number of false alarms is significantly reduced. This is because the system recognizes whether a person or a pile of earth is in the path of travel. This increases the efficiency of the work process.
This is made possible by the unique combination of a 3D PMD sensor and a 2D AI camera in an embedded system for area monitoring. By fusing sensor data from two different sensor types on a fully integrated and pre-calibrated platform, the system gains significantly in reliability. AI evaluation of the 2D camera data improves the quality of object recognition and enables differentiation between people and other obstacles. The new collision warning system thus works much better and more accurately than conventional solutions. The O3M AI smart sensor system generates a video from the data of 2D and 3D cameras that reproduces the environment and simultaneously displays information on the exact distances of all objects and people within a radius of up to 25 m in the image. O3M AI can differentiate at any time whether emergency braking, controlled stopping, or slowing down would be appropriate in the respective situation.
The digital, high-resolution Ethernet camera is equipped with a powerful processor with AI accelerator (NPU). This enables ultra-fast processing of AI algorithms for person recognition directly in the system. In addition to the Ethernet camera, a 3D PMD camera is integrated. Based on the PMD time-of-flight principle, this provides an exact distance measurement to the object in each pixel. The depth image from the 3D system and the video from the Ethernet camera can be combined or used individually, depending on the application.
The O3M AI's person and obstacle detection has been optimized for mobile work machines. The collision warning system can accurately distinguish a distant obstacle from a person in the immediate vicinity, even if that person is lying on the ground, wearing dark clothing, or is difficult for conventional systems to detect due to large tools or unusual body postures; and it can do so in bright sunlight as well as at dusk and up to a maximum range of 25 m with an accuracy of 10 cm.
The AI, hardware, and software were developed in Germany by ifm. The examples used to train the AI are taken from the typical working environment of mobile machines and include a wide variety of person types, equipment, and working positions. This results in a much greater gradation of possible alarms and warnings issued by the system. This is an important criterion when it comes to striking a balance between efficiency and safety.
The new combination sensor system for mobile machines is available in different versions. With integrated AI and person recognition, ifm markets it under the item number O3M372. It is part of a comprehensive ecosystem for mobile machines that ifm has built up over 35 years.