Researchers develop roadside warning device that can distinguish multiple objects and predict motion

According to foreign media reports, the Universities of Applied Sciences of Ulm and Heilbronn in Germany and their partners announced the development of the Salus system, which integrates radar, optical cameras, infrared sensors and machine learning neural networks. It can distinguish between pedestrians, cars, bicycles, motorcycles, deer, foxes, wild boars, etc. and predict their behavior.

(Image source: www.traffictechnologytoday.com)

The system can warn drivers and other road users to avoid traffic accidents. The data from the micro-Doppler radar is collected by Spectrum Instrumentation's PCIe digitizer card M2p.5926-x4, which meets the channel number and bit width required for this application.

该项目的领导者之一、乌尔姆应用科学大学Hubert Mantz教授表示，“我们项目的目标是在路边安装一些小型装置来探测危险，并在车辆接近时将其发送给车辆。此外，对于车内没有安装报警系统的道路使用者，将会接通路灯照亮危险区域，或照亮警告标志。SALUS项目将检测难以发现的危险并发出预警，最终显著提升道路安全系数。”

The technology demonstrator will be able to simultaneously measure data from radar, optical cameras and infrared cameras. In addition, the system can also integrate other sensors to measure pollution levels.

Hubert Mantz adds: “We are developing machine learning using neural networks that will enable the system to distinguish between a cyclist, a car or a deer, which goes far beyond pure motion detection. We are at a critical stage in the project, which is the classification of detected objects. Based on this capability, the system will be able to predict the movement of objects and thus potentially dangerous situations.”

The project envisages widespread deployment of these standalone units alongside roads across Germany. The need for such warning systems is greatest in rural areas where electricity supply and street lighting are inadequate, so they must be cheap and powered by solar energy.

At the same time, this also means that the communication system between the various units that make up the intelligent transportation infrastructure must have low power consumption characteristics. Therefore, the long-range wide area network (LoRaWAN) with a coverage range of 40 kilometers is the first choice for rural areas. It is characterized by low energy consumption and is based on unlicensed frequency bands, so the cost is low.