Concurrent in-situ flash X-ray radiography, Photon Doppler Velocimetry (pdv), and high-speed video using the ARL HIDRA provides microsecond-resolved information during ballistic impact experiments. Manual image analysis, although laborious and time intensive, provides characterization of dwell time, penetration velocity, and timing of break-out of the back surface of the target. To speed up the process and provide reproducible results image segmentation workflows were created using scripted image processing and machine learning in the MEDE Data Science Cloud (MEDE-DSC).
Scripted Image Processing (SIP) of HIDRA radiographs includes pre-processing to normalize contrast and perform film registration. Radiographs are split into separate images corresponding to the eight X-ray sources and alignment images are used to establish metrical relations between the different fields of view. Combined contrast enhancement (CLAHE) and band-pass filtering using Open CV algorithms provides a basis for projectile identification and feature extraction. While the scripts cut processing time from a day to approximately 15 minutes, they often fail to identify critical penetrator features in experiments with aberrant contrast or high penetration depth. To improve segmentation success rate, a convolutional neural network (CNN) with three hidden layers was created and trained by combining new shots with older imagery. Lack of training images remains one of the greatest challenges suggesting transfer learning using medical images with similar grayscale and complex objects may prove valuable.