Author Affiliations
1Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen 518000, China2Key Kaboratory of Optoelectronic Devices and System of Ministry of Education and Guangdong Province, College of Physics and Optoelectronic Engineering, Shenzhen University, Shenzhen 518060, Chinashow less
Fig. 1. Light path design of transmission(a) and reflection(b) and mathematical description (c) of terahertz single-pixel imaging based on modulator
Fig. 2. Schematic of the experimental setup for terahertz time-domain spectroscopy system
[46]. (a) The different colors correspond to different delay-line positions, these positions correspond to different time points in Fig.(b); (b) Measured THz time-domain signal; (c) The amplitude and phase of a signal in the frequency domain
Fig. 3. Terahertz single-pixel computational imaging based on metal pattern modulator. (a) PCB modulator
[50]; (b) Spinning metal disk modulator
[51] Fig. 4. Schematic diagram of terahertz single-pixel imaging structure based on metasurface modulator
[56]. (a) Spatial distributon of maximum differential absorption for Hadamard pattern; (b) Image reconstruction using 64 masks with each mask displayed for 22.4 ms, giving a total image acquisition time of 1.43 s; (c) Photograph of the object studied. (d) Consecutive tiles show reconstruction using 45 Hadamard masks
Fig. 5. (a) Schematic of terahertz single-pixel imaging system based on the passivated silicon wafer
[63]; (b) Comparison of reconstruction results between conventional high-resistive silicon and passivated silicon
Fig. 6. (a) Terahertz single-pixel imaging system and its imaging results based on nonlinear electro-optic crystals
[70]; (b) Terahertz single-pixel imaging system and its imaging results based on spintronic materials
[72] Fig. 7. Fourier single-pixel computational imaging for the terahertz regime
[81]. (a)-(d) Fourier spectrums at different sampling rates; (e)-(h) Inverse Fourier transform terahertz image
Fig. 8. (a) Network architecture of deep convolutional auto-encoder
[86]; (b) Design of part of optimized patterns; (c) Experimental structure and video of single-pixel imaging
Fig. 9. (a) DCAN architecture
[87]; (b) Qualitative and quantitative evaluation of conventional FSI and based on deep learning FSI
Fig. 10. Principle schematic of terahertz single-pixel computational imaging system at standoff distances and far-field imaging results
[92] Fig. 11. Schematic of terahertz single-pixel computational imaging construction and image reconstructed results
[94-95]. The imaging resolution is 154 μm, 100 μm and 9 μm when the intrinsic silicon thickness is (a) 400 μm, (b) 110 μm and (c) 6 μm, respectively
Fig. 12. Structural schematic of time-resolved nonlinear ghost imaging and hyperspectral reconstructed results
[97] Fig. 13. (a) Terahertz recognition experimental setup of target; (b) Perfor- mance evaluation (confusion matrices) of the object recognition method
[100] Type | Material | Modulation depth | Modulation rate | Reference | 注:*表示分辨率不可调谐的方法 | Metal* | PCB | 100% | - | [50]
| Spinning disk | 100% | 0.5 s | [51]
| Metasurface* | MMAs | −70 dB | 12 MHz | [56]
| Photo-induced
semiconductor
| Si | 20 dB | ~1.3 kHz | [59]
| GOS | 99% | - | [64]
| MEHPPV | 99% | 1.26 MHz | [67]
| VO2 | >75% | - | [68]
| Si-MOS | 15.3 dB | 1 GHz | [69]
| Source and
detector
| ZnTe | 100% | 1 kHz | [70]
| ZnTe | 100% | 1 kHz | [71]
| FM/NM | 100% | 22.4 ms | [72]
|
|
Table 1. Summary of terahertz wave mask technology
| Frequency range | Compressed sensing | Base scanning | Deep learning | Rt | Optical | - | <1 ms | <1 ms | THz | ~100 s
@64×64
| <1 ms | - | Sr | Optical | 2%
@256×256
| 1%
@256×256
| 4%
@128×128
| THz | 30%
@32×32
| 10%
@64×64
| - | Rq | Optical | MSE=0.04
@10%
| SNR=64.2 | PSNR=24 dB | THz | MSE=0.48
@30%
| SNR=6.2 | - |
|
Table 2. Comparison of technical indexes of single-pixel imaging algorithms