KEYWORDS: Non line of sight propagation, Target detection, Matrices, Imaging systems, Detector arrays, Singular value decomposition, Detection and tracking algorithms
Non-line-of-sight (NLOS) target localization technology achieves the localization of hidden targets by detecting and analyzing photons reflected or scattered by objects outside the direct line of sight. This approach has broad application prospects in fields such as autonomous driving and disaster rescue. This paper proposes a photon time-of-flight-based nonconfocal NLOS localization method, which employs a virtual wave algorithm for preliminary imaging of targets in complex scenes. Subsequently, the truncated singular value decomposition (TSVD) optimization method is utilized to extract the main features of the target solution. By reducing the computational time for redundant features and minimizing noise impact, this optimization achieves rapid and high-precision NLOS target localization. The experimental results indicate that the TSVD-optimized NLOS target localization method can accurately locate hidden targets. Within a target movement range of 50cm, the lateral positioning error is within 1cm and the axial positioning error is within 1.5cm, validating the high-precision target localization capability of this method in NLOS detection scenarios.
Due to the inhomogeneity of the medium and the scattering effect, phase distortion and intensity attenuation occur during the propagation of light, leading to the inability to obtain clear imaging of the target. Adaptive optics technology can compensate for the influence of scattering media on incident light waves by controlling the phase and amplitude distribution of the incident light. However, existing adaptive optics methods require prior knowledge of the target scene, which to some extent limits the application scenarios of this technology. To address this issue, this study constructs a no-reference image quality assessment system as a fitness metric. It iteratively generates the optimal compensating phase using a Genetic Algorithm(GA), enabling clear imaging of hidden targets in situations where the target scene information is unknown and there are no guide stars. Experimental results demonstrate that the employed no-reference image quality assessment system effectively constrains the optimization process. Specifically, the Energy gradient and Brenner gradient exhibit significant constraint effects in the early stages of evolution, showing a logarithmic improvement in imaging quality. The Tenengrad gradient performs best in the later convergence stage, achieving a peak signal-to-noise ratio(PSNR) of 14.34dB.
Imaging around corners which is also known as Non-Line of Sight (NLOS) imaging has been widely studied. In this paper, a method of NLOS imaging using steady state intensity detection signals is proposed. The proposed method only requires a consumer-level industrial camera, which greatly reduces the cost of detector parts required by traditional NLOS. Through the combination of the physical model and neural network, the proposed method can realize the online optimization of network parameters by an unsupervised strategy, and finally reconstruct the structure information of the original target. Therefore, the proposed method does not need to access the target plane and imaging plane to collect a large amount of paired data for neural network training in advance. Experimental results show that the proposed method can reconstruct targets with different structure types and can be applied to scenes with different intermediary surfaces. The robustness of this method increases the possibility of application in real scenarios, and this physical data driven joint optimization strategy will also inspire the thinking of other imaging methods in the field of computational imaging.
Based on the optical memory effect, many speckle-correlation methods have been proposed for imaging through scattering media. However, most of these methods directly take the laser with a rotating diffuser as a pseudo-thermal source without discussing its physical basis. To explore the effect of different optical coherence on object reconstruction, the principle that rotating diffuser can weaken the coherence is discussed and validated by experiments in this paper. Specifically, the degree of coherence is changed by controlling both the speed of the diffuser and the size of the illumination spot. And the speckle patterns collected under these conditions are reconstructed by Speckle-Correlation Algorithms (SCA). Through theoretical derivation and experiments, it is well demonstrated that the low coherence facilitates the recovery of the object. In addition, it is found that the stacked ground glass moving along a single direction can replace the rotating diffuser to achieve light source decoherence. The effect of different optical coherence on scattering imaging is explored in this paper, which provides a theoretical basis for selecting the light source in speckle-correlation imaging.
Strong scattering media bring difficulties to imaging in many fields such as medicine and astronomy. The deep learning method has a powerful fitting ability, which can be better applied in reconstructing the target behind the scattering medium. But the detail of the reconstructed target is often inaccurate. In this paper, the skip connection is added in the neural network to improve the accuracy of the reconstructed detail. This network can combine pixel-level information with high-level semantic information, and the information missed during the encoding process can be more involved in the final reconstruction process. 1100 handwritten characters are used as the targets hidden behind the ground glass. It is found that the quality of the reconstructed target is different when the skip connection is added to the network at different scales. The feature map visualization method is used to help us analyze the role of the skip connection. Meanwhile, PSNR (Peak Signal to Noise Ratio) is also used as an objective evaluation standard to evaluate the quality of reconstructed targets. According to subjective and objective evaluation criteria, conclusions can be drawn that the detail of targets can be better retained when the skip connection is added between the convolutional layer corresponding to the feature map of the size 64*64, and the average PSNR can be enhanced 1 dB compared with the network without skip connections. This work provides a reference for the fusion methods of different scale features in the computational optical imaging.
Recovering the object hidden in the disorganized speckle pattern generated through diffusive materials is an important topic as well as a difficult challenge. Existing speckle correlation imaging approaches generally use the speckle autocorrelation to extract the Fourier amplitude information of the target. Our goal here is to research the effects of the quality of the speckle autocorrelation on reconstructing targets via HIO-ER (hybrid input-output and the error reduction) algorithm. Specifically, a low-quality speckle pattern is preprocessed to estimate a high-quality autocorrelation. The PSNR of preprocessed autocorrelations could be increased from 5.88 dB to 24.08 dB. We also compare the differences between the preprocessed and unprocessed methods, and the reconstruction quality could be significantly improved than the later one. The result indicates that a high-quality speckle autocorrelation obtained after preprocessing helps to optimize reconstructing targets
The light scattering brings serious degradation for the object information. The conventional optical techniques cannot extract the relevant message on the object location in the scattering. In this paper, in the phase-space, the speckle characteristic with different depths has been analyzed and discussed. We utilize the phase-space-prior to locate the objects through a strong scattering medium with a learning method. Comparing with the single data-driven method, our scheme can help the deep neural network (DNN) to extract the depth information efficiently. The experimental results proved that our method is novel and technically correct with high locating accuracy. Our technique paves the way to a physical-informed DNN in locating and ranging objects through complex scattering media.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.