The National Transportation Safety Board (NTSB) has highlighted the importance of flight data monitoring as a tool to improve safety and efficiency of operations. In spite of the NTSB/s efforts, the participation rates of the rotorcraft industry in flight data monitoring programs are low due to the expensive cost of tools, anxiety of punitive action, etc. Since a video camera can be easily installed, is accessible, and inexpensively maintained, cockpit video recordings can replace the role of the Flight Data Recorder (FDR) in the absence of the FDR or can complement the FDR in a presence of the FDR. In this paper, we propose a DBSCAN (Density-Based Spatial Clustering of Applications with Noise) based image processing algorithm that accurately and efficiently estimates a helicopter's attitude such as the bank angle using cockpit image data. The proposed algorithm isolates a moving outside view from images of successive frames taken from the cockpit video camera using Average Motion Energy (AME) techniques, and estimates a helicopter's attitude using DBSCAN clustering. The performance of the proposed algorithm is tested and demonstrated by applying it to the cockpit image data.
Shin, S. and Hwang, I., “Helicopter Cockpit Video Data Analysis for Attitude Estimation using DBSCAN Clustering”, AIAA Infotech @ Aerospace, AIAA SciTech Forum, Jan 2016, https://doi.org/10.2514/6.2016-0920