Scannerless time-of-flight three-dimensional imaging, the successor to raster scanning three-dimensional imaging, relies on large-scale detector arrays to obtain high pixel resolution; however, manufacturing limitations lead to a bottleneck in imaging resolution. Here, we report a methodology for laser imaging using code division multiple access, which involves three key steps. Optical encoding is carried out for a pulsed laser to generate space-time encoded beams for projection. Optical multiplexing subdivides the backscattered light signals and multiplexes them to a downscaled avalanche-photodiode array. Subpixel decoding decodes the digitized encoded full waveforms and decomposes the features of all subpixels. Accordingly, we design a prototype based on a 64-order encoder and a 4-element avalanche-photodiode array and conduct an outdoor experiment. We demonstrate that the system is capable of obtaining 256 pixels per frame in push-broom mode and reconstruct a three-dimensional image with centimetre-level lateral resolution and range precision at a distance of ∼112 m.
10 December 2020 in IEEE Signal Processing Letters
The coprime array owns a sparse array structure and hence can effectively alleviate mutual coupling, while the holes existing in the difference co-array(DCA) greatly reduce the number of uniform degrees of freedom(uDOFs). In this paper, we propose a new coprime array structure called hole-free coprime array(HFCA), by carefully assembling the subarrays and the resulting HFCA can generate a hole-free DCA. Furthermore, we derive the optimal HFCA that can produce the largest hole-free DCA with a given number of sensors, which contributes to a substantial increase of uDOFs compared with the existing coprime array structures. In addition, the simulations are provided to demonstrate the superiority of HFCA in uDOFs, direction of arrival estimation performance, and spatial resolution.
Passive terahertz (THz) systems can perform the perspective imaging of objects concealed underneath clothing on the human body without a radiometric emitter. In this study, a linear-array passive THz imaging system was designed to realize highly efficient push-broom imaging. However, the acquired THz images are contaminated by severe striping and random noise, which greatly hampers object detection. Therefore, this paper aims to report a dedicated deep learning network for the automatic, accurate, and real-time detection of concealed objects in passive THz images. The main work of this research is threefold. First, a bilateral filter is integrated into a convolutional neural network (CNN) to generate a space-range tunable learning architecture. Second, a space-range grid combined with a geometric transformation matrix is designed to provide a core unit for multi-scale filtering and geometric (MSFG) augmentation. Finally, the one-stage YOLOv5 detector is improved to YOLOv5I by refining the granularity of its prediction units. The proposed approach was tested on onsite passive THz images acquired from various scenes, and the results demonstrate that the proposed MSFG augmentation method can significantly improve the detection accuracy of all the candidate CNNs on the passive THz images. Among them, the YOLOv5I-MSFG was found to exhibit the best performance via a comprehensive assessment of its real-time ability and detection accuracy; the time consumption per frame was found to be $\sim 53$ ms on the configured platform, and the accuracies of one-class and two-class detection were found to reach 93.49% and 90.76%, respectively.
The passive terahertz (THz) detection on objects concealed under clothing is quite challenging due to the heavy noise interference. In this paper, an active‐contour‐based self‐adaptive DBSCAN (AC‐SDBSCAN) detection algorithm is proposed. The core of AC‐SDBSCAN algorithm lies in that the object contours are first extracted by AC method in a noise scenario and then the statistical features of the contours are used to motivate a SDBSCAN to complete clustering without initialization. Benefiting from the strong robustness of AC and DBSCAN to the noise, the concealed objects in the noisy THz images can be detected accurately. Extensive simulations are verified on four passive THz image datasets. The results indicate that the AC method in our solution can achieve over 87% accuracy for contour extraction of passive THz images, while the classical methods achieve less than 77%; in addition, the SDBSCAN can achieve over 90% clustering accuracy without manual initialization which is significantly superior to the conventional DBSCAN. Eventually, the proposed method completes the object detection of passive THz images with a maximum recall of 90.38% and a maximum precision of 94%.
Existing deep learning methods cannot achieve satisfactory ship license plate (SLP) recognition due to the harsh marine environment, such as foggy weather, unstable ship state and small targets. Therefore, a convolutional recurrent neural network (CRNN)‐based method is proposed for accurate SLP image recognition. Overall, the suggested method improves a CRNN recognition model by SLP image enhancement and data augmentation. The SLP image enhancement employs dark channel prior and Hough transform line detector to address the fog/blurriness and tilt issues existing in SLP images. As separate and joint operations, the two enhancements contribute to data augmentation for CRNN recognition. Preprocessing algorithms, including adaptive histogram equalization and image edge padding, are used to improve and unify the enlarged dataset for augmenting the CRNN model. As a final step, correction of the CRNN recognition results is made according to the character rule of SLPs, using an edit‐distance algorithm to match against a pre‐established SLP dictionary. A variety of real SLP images were collected to build an SLP image dataset for verification. The experimental results indicate that our method can reach an SLP recognition accuracy of , which is significantly superior to other text‐based deep learning methods.