Fire Ignition Video Analysis Tool

OUSD (R&E) CRITICAL TECHNOLOGY AREA(S): Trusted AI and Autonomy; Advanced Computing and Software; Advanced Materials OBJECTIVE: Awardee(s) will develop a thermal video analysis tool capable of temporal and spatial discrimination of fragment flash and armor piercing incendiary (API) projectile function cloud in three dimensions using data from both high-speed visual and infrared camera systems. DESCRIPTION: Work has been underway for many years to understand the fire ignition phenomena in aircraft dry bays (void spaces) from the interaction of an impacting ballistic threat with a flammable fluid container such as a fuel tank. These threats include fragments produced by a warhead detonation, which cause a vaporific flash when impacting aircraft structural materials, or API projectiles, which result in release and ignition of incendiary material upon impact. The fragment or core of the projectile can penetrate the fluid container, releasing the fluid into a dry bay where the flash or incendiary functioning is occurring, and the overlap of these events results in an onboard fire. A fire ignition can start rapidly when the fluid is first spurting from the tank. During ballistic testing, many methods have been used to visualize the flash or function interaction with the leaking fuel. Both visual and infrared high-speed cameras have been used to capture the event as it unfolds. However, analysis of the event is difficult and subject to many variables, even when multiple cameras are employed from various angles and with multiple camera settings. Even with the presence of thermocouples and calibrated infrared cameras, it is very difficult to accurately capture temperature at precise locations over time. Both the timing and physical overlap in space of the flash/function and fuel are critical. Various data processing tools have been used to examine this issue, but identification of the size, location, volume, and relative intensity of the flash/function is very difficult. The presence of ignited material particles from the threat and spall from impacted aircraft surfaces makes characterization of the event even more difficult. This SBIR Phase I effort will focus on demonstrating a post-test processing tool for capturing both temporal and spatial characteristics of the flash/function cloud in relation to leaking fuel, so more informed models can be developed to predict fire ignition. The tool must help distinguish between the presence of the flash/function and the origin of fire ignition. This requires a more accurate characterization of the flash/function cloud, including the size, location, volume and intensity of areas of the cloud. The final tool must not only provide this characterization but must be able to track the presence and relative location of leaking fuel, particularly as it interacts with the cloud. This tracking must be possible in both low and bright light conditions. It must also provide a means for observing the event with both visual and infrared high-speed camera data and being able comparing and contrast these data. PHASE I: Awardee(s) will conduct a literature review and feasibility study to determine capability to characterize ballistic fragment flash and API projectile function size, location, volume, and relative intensity over time, while observing leaking fuel during dry bay fire testing. PHASE II: Awardee(s) will design, develop, and demonstrate a software tool capable of characterizing the temporal and spatial characteristics of ballistic fragment flash and API projectile function, an ability to distinguish these characteristics throughout the flash/function duration, identification of adjacent fuel leakage temporal and spatial characteristics, overlap of the two elements, and timing and location of fire ignition with minimal manual manipulation. PHASE III DUAL USE APPLICATIONS: Development of this tool will enable greater characterization of other ballistic event phenomena, such as warhead threat detonation and fragmentation, and hydrodynamic ram. It could also have commercial potential for forensic investigation of events recorded by video imagery. REFERENCES: Choi, J.; Han, T.; Lee, S.; Song, B. Deep learning-based small object detection. J. Inst. Electron. Inf. Eng. 2018, 55, 57 66. Kim, H.; Park, M.; Son, W.; Choi, H.; Park, S. Deep Learning based Object Detection and Distance Estimation using Mono Camera. J. Korean Inst. Intell.t Syst. 2018, 28, 201 20. KEYWORDS: ballistic threat; flash; function; incendiary; fire ignition; temporal; spatial