Autonomous car: Pedestrian detection in the dark with a fast infrared camera

Autonomous car: Pedestrian detection in the dark with a fast infrared camera

Assistance systems are intended to help relieve drivers and increase road safety – in some cases with automated functions. They are already a growing business area for developers and manufacturers. The future seems certain as regulations continue to get stricter and all players have high hopes for self-driving cars. Some systems have long been mandatory, and they do not always meet expectations. The young company Adasky has therefore further developed a sensor for one of these applications and will be presenting it for the first time at the IAA Mobility 2023 from September 5th to 10th in Munich.

Advertisement

Tightening of the rules after accidents

In the USA, the relevant regulations for monitoring the surroundings already explicitly apply to autonomous cars in everyday traffic. After accidents involving such vehicles, the National Highway Traffic Safety Administration (NHTSA) in the USA requested an emergency braking system at the end of May that can also protect pedestrians in poor visibility conditions. The authorities require reliable detection of human road users even at 0.2 lux, which is practically complete darkness. So far, 2000 lux has applied, which roughly corresponds to daylight.

With Adasky’s IR camera, the continuous detection of pedestrians should also be possible in the dark.

(Image: Adasky)

Adasky, developer and manufacturer of intelligent sensor technologies, therefore uses a long-wave infrared (LWIR) camera whose heat sensor can reliably detect objects with temperature differences of 50 millikelvin and more. In combination with a suitably developed algorithm, it should be possible to quickly and reliably distinguish people from other living beings and objects based on the temperature signature.

Regardless of lighting conditions

Regardless of the lighting conditions, the sensor should be able to detect objects up to a distance of 300 meters and classify living beings at a distance of more than 200 meters. The daylight cameras used today for Advanced Driver Assistance Systems (ADAS) in conjunction with low beam, on the other hand, only recognize an object from a distance of around 80 meters. But then the sensors are not yet able to determine whether it is a living being.

Calibration without flying blind

Advertisement

The camera uses multiple fields of view for different traffic scenarios. Also new is an ongoing, software-controlled calibration to the ambient heat. Conventional IR cameras, on the other hand, have to close a mechanical shutter for each calibration process, which can mean a considerable distance without detection depending on the driving speed. The data from the camera is used with the help of specially developed programs for object recognition and classification. The software is also capable of running clearance detection and even time-to-collision (TTC) calculations. According to Adasky, the new solution can be used to achieve both the required collision warning (FCW) and automatic emergency braking (AEB) in an emergency with a single camera.

Adasky has been developing and producing intelligent thermal sensor technologies since 2016 and currently has 80 employees, 60 of whom are engineers. The Israeli company takes a holistic approach to sensors and software in terms of road safety.

also read

Show moreShow less

(fpi)

Go to home page
#Autonomous #car #Pedestrian #detection #dark #fast #infrared #camera