Development of a system for airborne object detection to ensure safe operation of laser-optical ground stations

Kurzfassung

Satellite Laser Ranging (SLR) is a technology that could help to tackle the challenges arising from the quickly growing number of satellites, especially in lower earth orbits. To accelerate the expansion of the world-wide network of SLR stations, the DLR developed a transportable, low-cost SLR ground station, the miniSLR. However, the operation of powerful lasers in public airspace poses significant risks to aircraft flying nearby. To prevent hazardous illumination of aircraft, the miniSLR features an aircraft detection system. This thesis evaluates the existing aircraft detection system of the miniSLR and presents possible improvements. The transponder-based flight tracking and the performance of the neural net identifying aircraft in images of the thermal infrared camera installed on the miniSLR are critically reviewed. The analysis identifies the classifier as the main weakness of the detection system in terms of reliability of the detections and usability for laser ranging operation. In order to improve the detection system, a new dataset containing 1284 images across eight relevant categories is compiled. This dataset is then used to develop and evaluate four image analysis algorithms. A standout method, employing median filtering for background subtraction, is chosen as replacement for the neural classifier. It achieved a total of 99.8 % of correct classifications on the dataset, reducing the rate of misclassifications by more than a factor of six when compared to the neural net. This image processing technique is implemented in the miniSLR software and tested in operation. The performance impact of the image processing on the OOOS software is quantified and confirmed to be minimal, even at higher framerates, with a mean time to analyze one image of less than 8 ms. A preliminary analysis of detections recorded over a timespan of more than 24 hours suggested that in an operational scenario, at least 80 % of detections can be attributed to an aircraft.