Scholarlink

<p>In-line three-dimensional automated X-ray inspection (3D AXI) demands rapid reconstruction to meet strict cycle-time constraints, where the image quality needs only to be sufficient for reliable defect detection. Although beam-hardening correction (BHC) combined with the Simultaneous Algebraic Reconstruction Technique (SART) yields high-quality reconstructions, its computational cost is prohibitive for in-line applications. Incorporating global BHC into the Simultaneous Iterative Reconstruction Technique (SIRT) improves computational efficiency but suffers from averaging effects that reduce correction accuracy. To overcome these limitations, we propose a local BHC strategy specifically designed for integration with SIRT under circular computed laminography (CL) geometry. The method introduces spatially adaptive corrections that preserve ray-specific spectral information while maintaining the efficiency of SIRT. Validation with simulated and experimental printed circuit board assembly (PCBA) datasets containing ball grid array (BGA) solder joints, both with and without voids, demonstrated that SIRT-Local BHC effectively suppressed cupping artifacts, improved solder-void contrast, enhanced dimensional accuracy, and preserved boundary sharpness. Both simulation and experimental results consistently showed superior performance compared to SIRT-Global BHC. Moreover, a preliminary benchmark indicated that the additional computational cost of local BHC remains modest and compatible with in-line AXI cycle-time requirements. Importantly, the proposed correction does not require prior knowledge of material composition, material segmentation, or tabulated energy-dependent attenuation coefficient as inputs, and the correction strength is tuned directly from the acquired scan data under a fixed acquisition protocol. These findings highlight the potential of SIRT-Local BHC to improve imaging fidelity and defect detection reliability in high-speed in-line CL-based AXI, offering a practical solution for industrial applications where explicit material identity information is unavailable or impractical to obtain.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Iterative Reconstruction with Local Beam Hardening Correction for Reliable Computed Laminography without Prior Knowledge

  • 期刊论文

  • 发表日期：

    2026年3月7日

查看全文
  • Chia-Hao Chang,
  • Hsiang-Ning Wu,
  • Yu-Ching Ni

摘要

In-line three-dimensional automated X-ray inspection (3D AXI) demands rapid reconstruction to meet strict cycle-time constraints, where the image quality needs only to be sufficient for reliable defect detection. Although beam-hardening correction (BHC) combined with the Simultaneous Algebraic Reconstruction Technique (SART) yields high-quality reconstructions, its computational cost is prohibitive for in-line applications. Incorporating global BHC into the Simultaneous Iterative Reconstruction Technique (SIRT) improves computational efficiency but suffers from averaging effects that reduce correction accuracy. To overcome these limitations, we propose a local BHC strategy specifically designed for integration with SIRT under circular computed laminography (CL) geometry. The method introduces spatially adaptive corrections that preserve ray-specific spectral information while maintaining the efficiency of SIRT. Validation with simulated and experimental printed circuit board assembly (PCBA) datasets containing ball grid array (BGA) solder joints, both with and without voids, demonstrated that SIRT-Local BHC effectively suppressed cupping artifacts, improved solder-void contrast, enhanced dimensional accuracy, and preserved boundary sharpness. Both simulation and experimental results consistently showed superior performance compared to SIRT-Global BHC. Moreover, a preliminary benchmark indicated that the additional computational cost of local BHC remains modest and compatible with in-line AXI cycle-time requirements. Importantly, the proposed correction does not require prior knowledge of material composition, material segmentation, or tabulated energy-dependent attenuation coefficient as inputs, and the correction strength is tuned directly from the acquired scan data under a fixed acquisition protocol. These findings highlight the potential of SIRT-Local BHC to improve imaging fidelity and defect detection reliability in high-speed in-line CL-based AXI, offering a practical solution for industrial applications where explicit material identity information is unavailable or impractical to obtain.