We present a novel event-based quantitative flow visualization system, TrackAER, capable of continuously reconstructing, rendering and recording particle tracks in large test volumes without limitations on the measurement duration. Multiple event-based cameras are synchronized and calibrated to produce independent and asynchronous, yet temporally co-registered data streams of flow tracer positions. Subsequently, these data streams are merged into time-resolved three-dimensional particle tracks using photogrammetric techniques. Due to the operating principle of event cameras, the flow scenery is reduced to moving objects only, which effectively compresses the data stream at the camera source. In combination with an efficient data processing pipeline, the measurement system operates in real-time, reconstructing and rendering the investigated flow field without noticeable time lag. The data processing approach follows a “per-event” paradigm and enables the immediate observation and analysis of both, transient and long duration flow features. Specific issues resulting from event-based, frame-free processing are discussed as well as the advantages and limitations of event cameras. Exemplary results are provided to demonstrate the utility of the TrackAER system where Lagrangian particle track information is displayed in a virtual scene together with extracted quantitative information such as local flow velocities.
1 November 2021 in Optics and Lasers in Engineering
3D particle streak velocimetry (3D-PSV) and surface flow visualization using tufts both require the detection of curve segments, particle streaks or tufts, in images. We propose the use of deep learning based instance segmentation neural networks Mask region-based convolutional neural network (R-CNN) and Cascade Mask R-CNN, trained on fully synthetic data, to accurately identify, segment, and classify streaks and tufts. For 3D-PSV, we use the segmented masks and detected streak endpoints to volumetrically reconstruct flows even when the imaged streaks partly overlap or intersect. In addition, we use Mask R-CNN to segment images of tufts and classify the detected tufts according to their range of motion, thus automating the detection of regions of separated flow while at the same time providing accurate segmentation masks. Finally, we show a successful synthetic-to-real transfer by training only on synthetic data and successfully evaluating real data. The synthetic data generation is particularly suitable for the two presented applications, as the experimental images consist of simple geometric curves or a superposition of curves. Therefore, the proposed networks provide a general framework for instance detection, keypoint detection and classification that can be fine-tuned to the specific experimental application and imaging parameters using synthetic data.
We introduce a Convolutional Neural Network (CNN) that is specifically designed and trained to post-process recordings obtained by Background Oriented Schlieren (BOS), a popular technique to visualize compressible and convective flows. To reconstruct BOS image deformation, we devised a lightweight network () that has comparatively fewer parameters to train than the CNNs that have been previously proposed for optical flow. To train , we introduce a novel strategy based on the generation of synthetic images from random-irrotational deformation fields, which are intended to mimic those provided by real BOS recordings. This allows us to generate a large number of training examples at minimal computational cost. To assess the accuracy of the reconstructed displacements, we consider test cases consisting of synthetic images with sinusoidal displacement as well as images obtained in the experimental studies of a hot plume in air and a flow past and inside a heated hollow hemisphere. By comparing the reconstructed deformation fields using the or conventional post-processing techniques used in Direct Image Correlation () or conventional image cross-correlation, we show that is more accurate and robust in the synthetic test case. When applied to experimental BOS recordings, all methods provide similar and consistent deformation fields. As is capable of achieving a comparable or better accuracy at a fraction of the computational costs, it represents a valuable alternative to conventional post-processing techniques for BOS experiments.
The COVID-19 pandemic has emphasized the need for infection risk analysis and assessment of ventilation systems in indoor environments based on air quality criteria. In this context, simulations and direct measurements of CO2 concentrations as a proxy for exhaled air can help to shed light on potential aerosol pathways. While the former typically lack accurate boundary conditions as well as spatially and temporally resolved validation data, currently existing measurement systems often probe rooms in non-ideal, single locations. Addressing both of these issues, a large and flexible wireless array of 50 embedded sensor units is presented that provides indoor climate metrics with configurable spatial and temporal resolutions at a sensor response time of 20 s. Augmented by an anchorless self-localization capability, three-dimensional air quality maps are reconstructed up to a mean 3D Euclidean error of 0.21 m. Driven by resolution, ease of use, and fault tolerance requirements, the system has proven itself in day-to-day use at ETH Zurich, where topologically differing auditoria (at-grade, sloped) were investigated under real occupancy conditions. The corresponding results indicate significant spatial and temporal variations in the indoor climate rendering large sensor arrays essential for accurate room assessments. Even in well-ventilated auditoria, cleanout time constants exceeded 30 min.