摘要:
背景纹影（BOS）技术因其非接触、高时空分辨率的优点，已成为爆炸力学外场实验的关键测量手段。然而，强光干扰、产物散射及冲击波自身信号微弱、形态复杂等因素，使得从BOS图像中自动、精确地提取波阵面极具挑战。现有方法虽有进展，但波阵面提取精度及复杂环境处理仍亟待突破。为解决此难题，本文提出了一种结构感知加权变分光流法（SAW-VF），旨在对冲击波的高速瞬态位移场进行鲁棒量化。该方法的核心在于最小化一个精心设计的能量泛函：首先，其数据保真项融合了一阶光度约束与二阶Hessian矩阵不变性，显著增强了对冲击波线状局部几何特征的敏感度；其次，引入由归一化互相关（NCC）驱动的空间自适应加权机制，以动态抑制图像严重畸变区域的负面影响；再者，采用受Perona-Malik扩散理论启发的各向异性正则项，以有效保护冲击波锋锐的运动边界。整个优化过程嵌入由粗至精的高斯金字塔框架中，以应对大位移运动。在此基础上，本研究进一步提出一种物理模型驱动的波阵面拟合方法，通过最大内点集优化并结合冲击波动力学约束精确提取波阵面，并最终实现了非接触式的超压定量测量。在TNT爆炸试验中，本方法测量结果与压力传感器数据的相对误差范围为0.93%至9.85%，验证了该自动化系统在冲击波非侵入式超压测量方面的有效性与准确性。
Abstract:
Background-Oriented Schlieren (BOS) technique has become a key measurement method in external-field experiments of explosion mechanics, thanks to its advantages of non-contact and high spatiotemporal resolution. However, factors such as strong light interference, product scattering, weak intrinsic signals of shock waves, and their complex morphologies make the automatic and accurate extraction of wavefronts from BOS images extremely challenging. Although existing methods have made progress, the accuracy of wavefront extraction and the handling of complex environments still need urgent improvement.To address this problem, this paper proposes a Structure-Aware Weighted Variational Optical Flow (SAW-VF) method, aiming to robustly quantify the high-speed transient displacement field of shock waves. The core of this method lies in minimizing a well-designed energy functional: first, its data fidelity term incorporates the first-order photometric constraint and second-order Hessian matrix invariance, which significantly enhances the sensitivity to the linear local geometric features of shock waves; second, a spatially adaptive weighting mechanism driven by Normalized Cross-Correlation (NCC) is introduced to dynamically suppress the negative impact of severely distorted regions in images; third, an anisotropic regularization term inspired by the Perona-Malik diffusion theory is adopted to effectively preserve the sharp motion boundaries of shock waves. The entire optimization process is embedded in a coarse-to-fine Gaussian pyramid framework to cope with large-displacement motion.On this basis, this study further proposes a physics-model-driven wavefront fitting method. Through maximum inlier set optimization combined with shock wave dynamics constraints, the wavefront is accurately extracted, and finally, non-contact quantitative overpressure measurement is achieved. In TNT explosion tests, the relative error of the measurement results of this method compared with pressure sensor data ranges from 0.93% to 9.85%, which verifies the effectiveness and accuracy of this automated system in non-intrusive overpressure measurement of shock waves.
下载: