Passive millimeter-wave (PMMW) imaging is an ideal technique for concealed object detection in non-contact security inspection scenarios, and has been widely used in railway stations and airports. However, there are still several challenges that limit the accuracy of detection in PMMW images: low resolution, small objects and complex background interference. The existing deep learning-based methods mainly adopt two-stage architecture with convolutional neural networks (CNN) as the backbone for feature extraction, while the low speed of two-stage architecture and limited receptive field of CNN impede the further improvement of the intelligent inspection system. In this paper, we propose a one-stage anchor-free detector that combines the merits of CNN and transformer to solve these problems, and we avoid the tradeoff between accuracy and computational complexity that most hierarchical transformers make by constricting self-attention within pre-defined local windows. Specifically, we design a novel backbone to model low-level local and high-level global features at different scales via CNN and transformer. We also introduce object queries to guide the detector to perceive the concealed objects from the background noise, and these learnable queries are further utilized to form a full-size object-aware attention mechanism. Besides, to select the optimal positive samples for the anchor-free detector, we propose a novel label assignment strategy by employing Gaussian distribution to adaptively model the objects with various shapes. Experimental results on our self-developed PMMW imager demonstrate that the proposed method outperforms the state-of-the-art methods in terms of accuracy and speed.

Passive millimeter wave (PMMW) imaging has merits of non-radiation and good penetrability to most of clothes, hence it has been a reliable security technique for the detection of concealed objects. At present, deep learning based approaches have shown great advantage in automatic detection. However, low resolution and high background noise of PMMW images make the task tricky, especially for small objects. In this paper, we propose a transformer-based anchor-free detector with integration of local/global information and adaptive label assignment to address the aforementioned issues. Compared with the existing anchor-based methods adopted for PMMW image detection, our detector can further improve the efficiency and remove the handcraft anchor boxes. To be specific, we firstly employ hierarchical transformer architecture as backbone, which has the capacity to model long-range dependencies of the feature at different scales. We propose a new strategy that calculates self-attention within local region/global region in turn, providing detailed and global features of small objects. Secondly, we design a learnable position encoding module to obtain positional information between pixels. We propose an attention weighting module which enables the network to adaptively refine the features and distinguish positive and negative samples. Finally, we propose an adaptive label assignment strategy to dynamically optimize the number of positive samples used for detections. The proposed method is validated on our self-developed PMMW imager. The experimental results show that our method achieves better performance on accuracy and competitive speed compared with the state-of-the-art methods.

Heart rate measurement technology based on Doppler radar can monitor heart rate in real time without direct contact with the skin, which possesses the advantages of convenient operation, comfort and insensitivity to the skin state. However, due to the limitation of radar’s performance, heartbeat signals obtained by radar sensor do not have clear peak value like electrocardiograms (ECGs) and pulse waves, which affects its measurement accuracy. Therefore, this paper proposes a non-contact short-term heart rate detection system based on 120 GHz narrow beam frequency-modulated continuous-wave (FMCW) millimeter wave radar. On the basis of improving the signal-to-clutter ratio (SCR) of intermediate frequency (IF) signals with narrow beam, this system applies stochastic resonance algorithm to enhance the amplitude of heartbeat signals submerged in radar system noise, thus enhancing the signal-to-noise ratio (SNR) of heartbeat signals and improving the accuracy of heart rate detection. The short-term heart rate of 10 volunteers is measured at 3 s interval under the condition that the radar sensor was 1 m from the human chest. Experimental results show that compared with the peak count of the reference pulse wave sensor, the results of this system are highly consistent with those of photoplethysmograph (PPG) detection. Finally, those results of heart rate detection are also used to analyze heart rate variability (HRV).

A Ka-band 1024-channel passive millimeter wave (PMMW) imager with 1 GHz bandwidth called BHU-1024 has been developed by Beihang University for security screening. BHU-1024 uses linear phased array to obtain resolution in the horizontal direction and uses aperture synthesis to obtain resolution in the vertical direction. The non-ideal characteristics of the hardware cause the decrease of system sensitivity and always blur the reconstructed image, hampering applications where high resolution and accurate recognition are emphasized. Hence, effective calibration is a prerequisite to ensure the quality of the millimeter wave image. In this paper, the overall calibration approach is presented, focusing on the main errors relevant to the BHU-1024 instrument design: The basic correlator counts are preprocessed to correct for comparators offset and quadrature error. The residual offset is estimated by continuously sweeping the local oscillator (LO) phase. And for the baseline error, a novel and effective calibration approach, which makes use of an external noise source mounted on the metal frame, is proposed to tackle this problem. The results of simulation and measurement demonstrate the validity of the overall calibration approach. Finally, the actual calibrated data is used to reconstruct the millimeter wave image, which further validates the calibration process.

The design and measurement of a power detector circuit based on the 0.15-μm GaAs pHEMT technology and its usage in complex cross correlation measurement are presented in this article. The six-port technology is employed for the calculation of the cross correlation, and the design equations from the requirements of the correlation measurement are provided to guide the power detector design. The relation between the signal-to-noise ratio (SNR) of the correlation results and the load resistor is deduced, which can restrict the resistor value choice for an acceptable SNR. By properly choosing the load resistor and the biasing of the transistor, a power detector circuit is designed with a suppressed output dc voltage variation resulting from the statistic feature of monolithic microwave integrated circuit (MMIC) production. A fabricated detector chip has been mounted onto a printed circuit board and has been characterized. The measurement results show a flat responsivity over 3.5-8 GHz and a 3-nV/√Hz output noise voltage spectrum. The complex cross correlation measurement is implemented using two power detector modules with carefully chosen bias currents. Across the frequency range of 3.5-8 GHz, the measured correlation circles have an amplitude imbalance of 3.5 dB and an axial ratio from 0.85 to 1.15. With an input power of -5 dBm, the measured SNR of the correlator output is above 26 dB.

Passive millimeter-wave (PMMW) imager can detect concealed objects under clothing in a touch-free manner. Existing detectors exhibit promising performance on specific PMMW datasets. However, these detectors will suffer severe degradation due to the inevitable domain shift when the environment changes in real applications. To avoid huge manual efforts in preparing the target domain dataset, unsupervised domain adaption (UDA) is proposed to tackle this issue. UDA holds the assumption that the labeled source data and unlabeled target data are accessible during the adaptation, but it is not always true due to the privacy concern. In this paper, we aim to address this challenge by source-free domain adaption (SFDA), which migrates the source-trained model to target domain without the source data. Our method is built on the student-teacher framework, and combines the merits of pseudo-label self-training and adversarial learning. We also propose a novel spatio-temporal weighting strategy (STW). It mines the spatial self-correlation in single PMMW image and temporal cross-correlation among PMMW images for teacher to produce high-quality pseudo-labels. It guides the student to select strong positive pairs for contrastive learning. Besides, the adversarial learning is further employed for feature alignment between teacher and student at image and instance levels, which ensures the student to learn domain-invariant features. Extensive experiments show that our method achieves state-of-the-art performance. We further demonstrate that the proposed method can work in online settings. To the best of our knowledge, this is the first work to incorporate SFDA for concealed object detection in PMMW images.