A Novel Plenoptic Camera based Measurement System for the Analysis of Flight and Combustion Properties of Refuse-derived Fuel Particles Utilizing Tracking-by-Detection

In the past several decades, there has been a spate of interest in applying refuse-derived fuels (RDFs) in
industrial combustion processes, for instance, cement production. As a consequence of the economic
merit and the carbon-neutral characteristic, RDF owns a favorable application prospect. Nevertheless,
utilizing RDF for controllable and secure combustion is challenging since RDF is composed of various
waste fractions with complex shapes resulting in relatively complicated flight and combustion behaviors.
The undertaken research presents a novel plenoptic camera based measurement system to determine
the properties of RDF particles using image processing approaches. At first, the particles are captured
by a plenoptic camera, which is able to provide information in 2D gray value images and 3D point
clouds, i.e., each spatially captured pixel contains both gray value as unsigned integer number stored
with 16 bit and spatial coordinate in mm. Based on the information, the particles can be detected by 2D
gray value based algorithms and 3D clustering approaches. Owing to the considerable fluctuation of
the obtained point clouds, 3D clustering approaches perform inferiorly. ... mehrCompared to the 3D clustering
approaches, the 2D gray value based algorithms also exhibits particular deficiencies despite their better
accuracy performance. Under the circumstance, the study proposes an innovative combined detection
approach that combines the detection results of a 2D gray value based algorithm and a 3D clustering
method aiming at making the most of the acquired 2D and 3D information to deal with the fluctuations
of the measuring system. Subsequently, the particles are tracked by the linear Kalman filter, with
the 2.5D global nearest neighbor (GNN) and the joint probabilistic data association (JPDA) approach,
respectively. By introducing the depth information in gating, the tracking approach is extended to
2.5D. As a consequence of several inaccurate detection results caused by the measuring system, the
initial tracking results contain faulty and incomplete tracklets that entail a post-processing process.
Therefore, in the study, a post-processing framework based merely on particle motion similarity is
developed, which benefits a precise tracking performance by eliminating faulty tracklets, deleting
outliers, connecting tracklets, and fusing trajectories. The thereby obtained 2D particle trajectories
are converted into spatial particle traces in accordance with the provided 3D information by the
plenoptic camera. Subsequently, the trajectories are estimated by polynomials independently in three
spatial directions for the sake of compensating fluctuations existing in the trajectories. To conduct
a quantitative assessment of the proposed approaches, manually labeled ground truth datasets for
particle detection and tracking are generated. The presented methods in the study are proven to deliver
satisfactory performances with respect to several measurements, for instance, precision, recall, and
F1-score.
The thereby obtained complete spatial fuel trajectories enable the analysis of the flight and combustion
behaviors of various fuel fractions. For the flight properties, the flight duration and space-sliced particle
velocity in the depth direction are elaborated. The combustion behavior refers to the ignition time. As
revealed by the study, wood chips and PE granules move more rapidly than the two paper materials.
While paper shreds and confetti could be ignited during their flights, PE granules and wood chips land
before ignition. The acquired statements demonstrate an excellent agreement with the computational
simulations, which reversely prove the availability and applicability of the developed measurement
system. This new measurement system and the provided experimental results can benefit a better
understanding of the RDF’s combustion for future research.