Image processing in airborne applications using multicore embedded computers

摘要

More autonomous flying platforms will have to rely on on-board image processing performed in real time due more sophisticated applications like more autonomous landing or collision sensing, object detection, recognition and identification. This work is an effort to investigate whether modern COTS multicore embedded computers are suitable to perform state-of-the-art image processing applications, such as object recognition, on-board an aircraft. This paper presents requirements of avionics, challenges of implementing such algorithms, the characteristics of possible hardware platforms, and their programing model considerations that influence the performance of such applications. The paper presents an experimental implementation of core algorithm blocks on multicore platforms from Freescale. It shows that modern multi-core processors still need a performance boost for demanding real-time image processing.