Perception is a major challenge for autonomous mobile systems. The detection of objects in the immediate environment plays an important role here. To overcome this challenge, the use of optical systems such as cameras presents a suitable solution. With the generated data, the respective relevant objects can be detected with the help of AI-based methods. One possible example of such a system is the detection of traffic lights and their states in the autonomous vehicle. While the position of such a system is often stored in highly accurate maps, the detection of its state relies on a reliably functioning perception of the environment.
The relevant environmental objects should be reliably detected. For the aforementioned example of traffic light detection, this means reliably detecting all traffic lights in the path of the autonomous vehicle and avoiding false positive detections.
With the help of AI algorithms, the information contained in the image data is to be extracted and made technically usable for the respective application purpose.
Evaluation of the developed algorithms on problem-specific data sets.