Where people and heavy machinery move in the same environment, accidents are unfortunately still part of everyday life. In mining, on construction sites, in logistics and elsewhere, vehicle drivers navigate areas that are difficult to see with machines that have very large blind spots. Companies are now increasingly looking for sensory aids to ensure occupational safety. Various solutions are in competition, based on radar, lidar, ultrasound, RFID or cameras. Retenua has presented a fully integrated system for mounting on vehicles that detects reflectors on high-visibility vests. It is based on an infrared camera and can trigger alarms and braking processes directly. Vision Components contributes the intelligent stereo camera. This embedded vision system with a powerful integrated processor captures the images and evaluates them itself. This means that no peripheral devices are required for image processing and the solution is extremely robust, both mechanically and functionally.

NIR 3D people detection

Retenua has patented a method for detecting retroreflectors - the reflective strips on high-visibility vests and protective clothing - and developed the Emitrace driver assistance system on this basis. This consists of a sensor and computer unit that is usually mounted at the rear of industrial vehicles. An optoacoustic signaling device is located in the driver's cab. No other devices are required. The integrated embedded vision system detects reflectors, calculates the 3D position and speed of people from the stereo images and triggers an alarm in the cab if there is a risk of an accident. It is even possible to link Emitrace directly into the vehicle control system so that it can automatically initiate braking.

The modules are extremely robust and can be used in a wide temperature range from -30 to 60 °C. The sensor unit has IP69K protection at the front and IP67 at the cable connection and is easy to clean. The viewing windows are made of hardened, scratch-resistant glass. The powerful NIR (near infrared) flash allows short exposure times, minimizing the influence of infrared radiation in the environment. The flash does not disturb people in the vicinity, but ensures stable detection performance in any lighting. Emitrace is suitable for indoor and outdoor use, day and night, and could become standard in mining applications in particular.

Demand-driven development

Rafael Mosberger and Henrik Andreasson conducted research at the University of Örebro in Sweden on detecting people with industrial vehicles. There they developed their new method for detecting retroreflectors with an infrared camera. The work was supported by both EU funding and industrial companies, which led the two scientists to found Retenua in 2015. The start-up provided several vehicle manufacturers with an early version of Emitrace as a developer kit. There was a lively exchange with their design departments in order to understand their needs and arrive at the current plug-and-play standard product with preset parameters.

Emitrace is designed for harsh industrial environments. © Retenua

Emitrace is designed for easy retrofitting on various mobile machines and for maintenance-free operation and has already been successfully tested in the field. Retenua works together with sales partners for installation and commissioning. These include forklift truck manufacturers, rental companies and service providers. Retenua's C++ vision library and the API provided allow flexible integration into different software environments. This means that raw data and image processing results can also be transferred to other systems for further processing or documentation.

Comparison with lidar, RFID and camera

Emitrace increases personal safety in the vicinity of heavy machinery. Lidar, radar and ultrasound, on the other hand, cannot reliably distinguish between people and objects. In industrial environments, which are characterized by narrow paths, they would constantly issue false alarms. Some products are based on RFID. This is a practical alternative to lidar, for example, and has the advantage that the transponders are detected even if they are obscured by obstacles. However, considerable costs can be incurred because staff and the vehicle fleet have to be equipped with transponders and readers.

With the exception of Emitrace, all existing camera-based products are trained to recognize people based on their silhouette. When walking and standing, this is easy to solve with algorithms. Other postures such as kneeling, sitting or lying down still overtax computing power. Reliable algorithms sometimes need several seconds to process a single image. It will be several years before this capability can also be implemented in real time. As Emitrace is trained exclusively on retroreflectors, body posture is no obstacle for the system.