Hardware Support Vector Machine (SVM) for satellite on-board applications

Since their introduction in 1995, Support Vector Machines (SVM) have shown that classification by this relatively recent machine learning tool can be more accurate than popular contemporary techniques such as neural networks and decision trees, hence causing it to find its way quickly to various applications in engineering, economy and statistics. Despite their possible advantages, SVM use in space applications is still very limited for several reasons including low technology maturity and high computational demand. This paper proposes overcoming the computational demand hurdle through a hardware friendly implementation of SVM for satellite onboard applications using FPGAs. The evaluation of the proposed system shows excellent classification accuracy, low device utilization and acceptable speed for satellite onboard applications. The results shown in this paper opens the door for further exploration of various possible onboard applications including on-board image analysis, compression and autonomy.