GPU-accelerated fast near-field millimeter-wave imaging technique for planar scanning of one-dimensional nonuniform sparse arrays

• Published in: Journal of electronic imaging Vol. 34; no. 4; p. 043035
• Main Authors: An, Deyue, Hu, Shiping, Shi, Yang, Xu, Huimin, Zhu, Lian, Yao, Qingfeng
• Format: Journal Article
• Language: English
• Published: Society of Photo-Optical Instrumentation Engineers 07.08.2025
• Subjects: planar scanning millimeter wave imaging; GPU-accelerated; range migration and back-projection algorithm; millimeter wave imaging; synthetic aperture imaging
• ISSN: 1017-9909; 1560-229X
• DOI: 10.1117/1.JEI.34.4.043035

To reduce hardware costs, our previous work proposed a near-field millimeter-wave imaging technique for one-dimensional nonuniform sparse array planar scanning systems, featuring the innovative range migration back-projection algorithm (RMBPA). This hybrid algorithm applies the range migration algorithm in uniformly sampled vertical and range dimensions while employing the back-projection algorithm in nonuniformly sampled horizontal dimensions, achieving high-quality imaging for sparse array configurations. However, the CPU-based RMBPA implementation required 49 s for image reconstruction, which far exceeded the practical engineering requirements. This paper introduces a GPU-accelerated solution, which achieves a reconstruction time of 0.13 s, representing a 370 times speedup compared with the original method while preserving the same level of performance. We first detail the implementation methodology for GPU-accelerated image reconstruction in one-dimensional sparse array systems, employing the cuFFT library for fast Fourier transform acceleration and specifically optimizing the computationally intensive interpolation and back-projection operations. Resolution tests and human subject experiments then validate the technique’s capabilities. The key results indicate that the method achieves a 5 mm spatial resolution, which is on par with the CPU-based results (MAE <1.7  dB) while realizing a significant reduction in reconstruction time. In addition, security-screening trials have successfully detected a variety of concealed threats, such as ceramic knives and mock firearms, under clothing, thereby verifying the practical applicability of the proposed technique. The proposed solution delivers an efficient, high-precision, and cost-effective imaging approach for millimeter-wave security systems, with significant engineering application value.