ИСТИНА

Modern digital high-speed cameras and sensors generate large amount of digital information. The need arises to automate the processing of large arrays of images and video recordings obtained during the visualization of gas-dynamic flows. In the present study, we applied shadowgraph and schlieren techniques to visualize gas flows. We developed 2 programs for automating the processing of shadowgraph and schlieren images: one is based on computer vision methods (background image subtraction, noise removal, edge detection, Hough transform); the other is based on machine learning (convolutional neural network). We applied them to detect and track normal shock waves, oblique shocks, pseudo-shocks and tracer particles in the flow. In the present study we tested our software on shadowgraph and schlieren images of gas flow in the shock tube. We obtained videos using high-speed camera with the recording frame rate up to 150 000 frames/s. Using the developed software, flow in shock tube was automatically quantitatively investigated, including the initial shock wave, reflected pseudo-shocks (shock train) and gas flow following the initial shock wave. Also, we studied oblique shock from point obstacle placed on the bottom wall of the shock tube. Oblique shock angle was measured automatically by our software. Gas flow velocity behind the initial shock wave was estimated by the oblique shock angle and by the tracer particles velocity in the flow.