Synergistic Event-SVE Imaging for Quantitative Propellant Combustion Diagnostics

Real-time monitoring of high-energy propellant combustion is difficult. Extreme high dynamic range (HDR), microsecond-scale particle motion, and heavy smoke often occur together. These conditions drive saturation, motion blur, and unstable particle extraction in conventional imaging. We present a closed-loop Event–SVE measurement system that couples a spatially variant exposure (SVE) camera with a stereo pair of neuromorphic event cameras. The SVE branch produces HDR maps with an explicit smoke-aware fusion strategy. A multi-cue smoke-likelihood map is used to separate particle emission from smoke scattering, yielding calibrated intensity maps for downstream analysis. The resulting HDR maps also provide the absolute-intensity reference missing in event cameras. This reference is used to suppress smoke-driven event artifacts and to improve particle-state discrimination. Based on the cleaned event observations, a stereo event-based 3D pipeline estimates separation height and equivalent particle size through feature extraction and triangulation (maximum calibration error 0.56%). Experiments on boron-based propellants show multimodal equivalent-radius statistics. The system also captures fast separation transients that are difficult to observe with conventional sensors. Overall, the proposed framework provides a practical, calibration-consistent route to microsecond-resolved 3D combustion measurement under smoke-obscured HDR conditions.

💡 Research Summary

The paper tackles three intertwined challenges that arise during high‑energy solid propellant combustion: extreme high dynamic range (HDR) exceeding 120 dB, microsecond‑scale particle motion, and dense, time‑varying smoke that obscures visual information. Conventional high‑speed cameras cannot simultaneously provide sufficient dynamic range, temporal resolution, and robustness to smoke; laser‑based diagnostics suffer from attenuation and multiple‑scattering in optically thick plumes; and single‑sensor approaches inevitably trade off one requirement for another. To overcome these limitations, the authors propose a tightly coupled measurement system that fuses a spatially variant exposure (SVE) camera with a stereo pair of neuromorphic event cameras.

System Architecture and Calibration
The SVE camera uses a 2 × 2 macro‑pixel mask containing four neutral‑density filters, enabling four interleaved exposures to be captured in a single snapshot. This one‑shot HDR acquisition eliminates motion‑induced ghosting that plagues multi‑frame exposure fusion. The two event cameras operate asynchronously, reporting logarithmic brightness changes (events) with microsecond timestamps and a dynamic range >120 dB. Precise spatial and temporal alignment is achieved through hardware triggering, a special timestamped event injected into each stream (global time reference), and a shared calibration target visible to all sensors. Intrinsic parameters are estimated via Zhang’s method; extrinsic parameters define the left event camera as the world frame, and a rigid transform aligns the SVE frame to the stereo event system. Synchronization jitter is measured at ≤12 µs, which translates to sub‑pixel displacement for the observed particle velocities.

Smoke‑Aware HDR Fusion
Because combustion simultaneously produces intense point‑source radiation from particles and dense volumetric smoke, conventional multi‑exposure HDR (MEF) algorithms mix the two signals, degrading particle detail. The authors introduce a probabilistic smoke‑likelihood map (F(\mathbf{x})) built from four complementary cues extracted across the four SVE exposures: (1) brightness deviation (B_I) that normalizes intensity differences while suppressing low‑SNR noise, (2) Weber contrast (W_C) emphasizing edge strength relative to local brightness, (3) a dark‑bright channel contrast (C_F) derived from atmospheric scattering models to highlight optically thick regions, and (4) normalized inter‑exposure variance (V) to detect saturated or highly non‑linear pixels. A weighted sum (\alpha B_I + \beta W_C + \gamma C_F + \sigma V) (with empirically set weights (\alpha=0.1, \beta=0.4, \gamma=0.2, \sigma=0.3)) yields a scalar field where higher values indicate stronger smoke influence. This map guides the HDR fusion process: pixels with high smoke likelihood are treated more conservatively, preserving particle edges while attenuating smoke‑induced attenuation. The resulting HDR frames provide absolute radiance values that are otherwise missing from event data.

Intensity‑Guided Event Processing
Event cameras excel at capturing rapid brightness changes but lack an absolute intensity reference, making it difficult to distinguish genuine particle events from spurious events generated by turbulent smoke. By projecting the SVE‑derived HDR intensity onto the event image planes, the system obtains a per‑pixel intensity prior. Events occurring in high‑smoke‑likelihood regions are filtered out, while those in low‑likelihood regions are retained. Additionally, the absolute intensity values are used to normalize event polarity and rate, further suppressing smoke‑driven artifacts. The cleaned event streams retain precise timing and spatial information about particle emission, enabling reliable downstream analysis.

Stereo Event‑Based 3‑D Reconstruction
From the cleaned event streams, feature points (edges, corners) are extracted using event‑based corner detectors. Because the two event cameras are synchronized and calibrated, corresponding features can be triangulated to recover 3‑D positions. The pipeline estimates two key combustion metrics: (i) separation height—the distance from the burning surface to the point where particles detach, and (ii) equivalent particle radius, assuming roughly spherical particles. Calibration experiments demonstrate a maximum geometric error of 0.56 % across the measurement volume, confirming the method’s quantitative accuracy. The system can thus resolve particle trajectories and size distributions at microsecond temporal resolution.

Experimental Validation
The authors validate the framework on boron‑based solid propellant combustion, a scenario known for producing coral‑like boron agglomerates and dense smoke. HDR maps successfully separate bright particle emission from smoke scattering, preserving fine particle boundaries. The intensity‑guided event processing eliminates smoke‑induced false events, leading to clean particle tracks. Stereo event triangulation captures fast separation transients that are invisible to conventional high‑speed cameras due to saturation or motion blur. Statistical analysis of the reconstructed particle sizes reveals multimodal distributions, reflecting the known stages of accumulation, coalescence, growth, and detachment of boron particles. The system also records rapid height changes during detachment, providing insight into the dynamics of particle ejection.

Contributions and Limitations
Key contributions include: (1) a combustion‑specific HDR fusion algorithm driven by a multi‑cue smoke‑likelihood map, (2) an intensity‑guided event filtering strategy that leverages HDR frames to suppress smoke artifacts, and (3) a stereo event‑based 3‑D measurement pipeline delivering microsecond‑resolved particle height and size estimates with sub‑percent calibration error. Limitations noted by the authors involve the empirical selection of smoke‑likelihood weights, which may need retuning for different propellant chemistries or lighting conditions, and potential event‑camera saturation in extremely dense particle clouds. Future work is suggested on automatic parameter optimization, adaptive weighting, and hardware designs to mitigate event saturation.

Overall Impact
By integrating SVE imaging and neuromorphic event sensing in a closed‑loop architecture, the paper delivers a practical, calibration‑consistent solution for real‑time, quantitative combustion diagnostics under the most demanding optical conditions. The ability to obtain absolute radiance, suppress smoke‑induced noise, and reconstruct 3‑D particle metrics at microsecond scales opens new avenues for propellant formulation optimization, instability suppression, and high‑performance rocket engine development.