Performance analysis of vision AGL light spots detection using embedded devices
Kacper Podbucki, Tomasz Marciniak · 2025
The article presents an analysis of the issue concerning the use of embedded devices for the task of detecting the light points of airport AGL lamps. This type of detection is an important preliminary stage of the measuring process for testing the photometric parameters of in-pavement AGL lamps. The algorithm developed during previous studies was tested in two versions: supporting calculations on the CPU and using CUDA technology and performing calculations on the GPU. An experiment was conducted that involved the processing of video sequences from airport areas for the detection of in-pavement AGL lamps in various weather conditions and times of day. The efficiency of the programs was checked by analyzing the average number of frames per second achieved by the program. The software was compared on single-board microcomputers of the Raspberry Pi family and modules of the NVIDIA Jetson family, for which changes in power modes were also taken into account.