| A Stereoscopic Imaging Velocimetry (SIV) that can provide three-dimensional (3-D) three-component (3-C) gross-field flow measurement has been developed. The SIV is an optical non-intrusive technique based on a two CCD cameras system, which simultaneously captures images of tracer particles seeded in a flow, from two different vantage points. When a flow field with a particle seeding of high density is imaged on CCDs, many individual particle images, that are equivalent data points, are likely to be lost in data processing. The data loss and erroneous detection mostly occur in the process involving particle overlap decomposition of superimposed particle images and particle tracking across image frames. To maximize data point recovery and to enhance measurement reliability, artificial neural networks are implemented in the two phases of the SIV technique. For particle overlap decomposition, a Back-Propagation neural network is utilized to take advantage of its ability in nonlinear classification and pattern recognition. A Hopfield neural network and a stochastic neural network are employed to provide significantly reliable particle tracking solutions by using the massive optimizing processing power of the neural networks. Both computational and experimental results are presented. The results have proved that the SIV to be a viable candidate to reliably measure asymmetric unsteady flow fields. The SIV is advantageous in system simplicity for building compact hardware and in algorithm efficiency for continual near-real-time velocity field monitoring. It is believed that the SIV technique can provide appropriate means for near-real-time 3-D 3-C flow diagnostics of a restricted volume with reasonable measurement accuracy. |
