In autonomous driving, redundant perception is critical for robust environmental understanding, especially under adverse conditions where traditional sensors like RGB-D fail. This paper introduces a novel Point Transformer for mmWave-radar-based pedestrian 3D reconstruction (mmPPT), addressing the challenges of sparse, noisy 4D mmWave radar point clouds. mmPPT employs a novel Hierarchical Serialization Strategy that combines Hybrid Local Serialization to preserve fine-grained body details via adaptive space-filling curves and Anchor-based Global Serialization to model long-range structural relationships. Additionally, it integrates temporal information and Doppler velocity features to leverage radar’s unique capabilities, while a Bone Length Loss is adopted to enforce geometric constraints on joint position prediction. Experiments on 4D radar dataset demonstrate that mmPPT outperforms state-of-the-art methods, achieving the lowest total reconstruction error (12.76 cm with multi-frame input) and superior robustness in harsh environments like smoke, occlusion, and poor lighting scenarios where RGB-D sensors fail. Ablation studies validate the effectiveness of the Hybrid Local Serialization Strategy and the Anchor-based Global Serialization Strategy, highlighting mmPPT’s potential to enhance redundant perception in autonomous driving. The source code will be available at https://github.com/yanglijie-dev/mmPPT.

A microwave photonic radar with ISAR imaging resolution better than 2.0 cm is proposed and experimentally demonstrated, in addition, the phase variation of 0.36 rad within 10 seconds in receiver is achieved.

The implementation of IM-DD optical communication systems requires low cost, power consumption and latency, making low complexity and high real-time deployment potential digital signal processing (DSP) a critical component. The clock synchronization algorithm plays a crucial role in ensuring real-time deployment by eliminating sampling clock deviation at the transceiver and optimizing the signal sampling position. In this context, we proposed an enhanced clock synchronization scheme based on the classic Gardner algorithm, which effectively addresses the widespread timing jitter issue in the synchronization loop while maintaining computational efficiency. Through comparative experiments of multiple single-lane 112Gb/s PAM-4 IM-DD transmissions, we demonstrated that our clock synchronization scheme is capable of effectively compensating for the clock offset in a high-speed 100Gb/s+ IM-DD transmission system. Our experimental results show that the algorithm achieves a correction accuracy of 1 part per million level, which is comparable to the receiver sampling frequency in the transmission (GHz). Moreover, the importance of clock synchronization for IM-DD systems is also highlighted by comparing signal quality with and without the algorithm. Without an equalizer to compensate for signal degradation after using a clock synchronization algorithm, the constellation can still be obtained clearly. If not perform the clock synchronization, a clear constellation cannot be obtained even with an equalizer. This is manifested in the signal-to-noise ratio, which our clock synchronization algorithm can bring the improvement of more than 1 dB. Additionally, the bit-error-ratio can be improved by about 5 numerical levels under specific modulation amplitude.

77 GHz radar has become a promising approach to enhance automotive safety by quickly detecting and identifying targets around the vehicle, especially in harsh weather conditions. This requires 77 GHz radars to provide environmental imaging with high resolution and reliability. However, radar images are easily blurred by sidelobes and background noises, which makes it difficult to extract real target information. In this paper, a sidelobe suppression algorithm based on the point spread function (PSF) and complex-valued neural network has been proposed to discriminate and suppress unwanted sidelobes while maintaining mainlobes referring to targets. To overcome the scarcity of real-world 77 GHz multiple-input multiple-output (MIMO) radar datasets, this paper derives the formula of PSF for 77 GHz MIMO radars in detail and exploits the PSF to generate simulated datasets for training. In addition, to be compatible with the complex-valued radar datasets, a customized neural network model has also been established in this paper. The well-trained neural network is further adopted to suppress sidelobes on real-world radar images. Comprehensive simulations and measurements have proved the superior performance of the proposed method, especially in cases with low signal-to-noise ratio (SNR), large channel mismatches, small target Radar Cross-Section (RCS) and large observation angle.

Millimeter-wave (mmWave) imaging is becoming more necessary in security inspection as it penetrates fabrics, leathers, and other materials and is harmless at appropriate power levels. Currently, better restoration of the original scene from the radar echoes with interference is prime attention of researchers. However, with the short wavelength of mmWave, a slight positional bias can dramatically affect the focusing performance of imaging. Notably, mechanical scanning generally obtains a large virtual aperture in near-field mmWave high-resolution imaging applications. However, the mismatch bias between automated scanning and radar parameters accumulates and aggravates the image defocus. This paper studies the scanning bias of the 77 GHz mmWave imaging system based on the 3DRIED dataset, analyze the mismatch bias in both the beginning moment and velocity, and gives the corresponding model. Meanwhile, referring to the range alignment method of ISAR, this paper presents a scanning data alignment algorithm only using MISO radar echoes. The algorithm calculates the relative bias of two adjacent scanning by comparing the data of horizontal scanning to get the spatial distribution of all sampling points. By comparing the results before and after data alignment, the algorithm shows that as a pre-processing method, it can effectively improve the focusing performance of the high-resolution scene reconstruction.

The invention discloses an underwater cross-medium communication and water surface target detection and tracking integrated system. The system carries out detection and tracking on a water surface target according to the echo intensity while sensing the weak vibration of a water surface through the echo phase change of a frequency-modulated continuous wave radar for cross-medium communication. Thesystem comprises an underwater acoustic transducer used for transmitting modulated communication underwater acoustic signals underwater, and a millimeter wave radar system used for detecting a watersurface target in the air. Sound waves emitted by the underwater acoustic transducer can generate weak vibration on the water surface, modulation signals carried by the sound waves can be extracted byutilizing the weak change of the phase of echo signals reflected by the radar water surface, and corresponding communication signals can be obtained through modulation processing. While underwater cross-medium communication is carried out, the echo intensity of the radar in other areas is utilized to detect and track the water surface target, and the method is suitable for the field of radar communication integration.

A new cascading multiple input multiple output (MIMO) radar at V-band with 192 virtual elements is developed. The experiment for imaging indoor and outfield is performed with an angular resolution of 0.6°. In addition, sparse array with 108 virtual elements based on this platform is tested with genetic algorithm, the imaging results show that the sparse array imaging with a sparse rate of 55.1% agree well with full array.

The invention discloses a solid-state laser radar detection method based on a cyclic frequency shift ring, and the method comprises the steps of: transmitting a narrow-band swept-frequency optical signal to the cyclic frequency shift ring to obtain a broadband swept-frequency optical signal, dividing the broadband swept-frequency optical signal into two paths, taking one path as a reference optical signal, and transmitting the other path to a dispersion unit through an optical collimating mirror; controlling different-frequency swept-frequency sub-signals of the broadband swept-frequency signal to sequentially point to different directions in a free space by the dispersion unit to obtain a series of detection optical signals pointing to different directions; reflecting the detection lightsignalback to the dispersion unit after encountering a target, receiving the detection optical signals by the optical collimating mirror, and performing combination with the reference optical signal to form a path of to-be-detected optical signals; and conducting photoelectric conversion and signal acquisition on the to-be-detected optical signals, then acquiring target space distribution information based on a signal processing algorithm. The invention also discloses a solid-state laser radar detection system based on the cyclic frequency shift ring, and the system can achieve the high-precision measurement of a target angle and distance information at the same time through the optical cyclic frequency shift and optical wavelength dispersion technologies.

With the development of millimeter-wave radar technology, synthetic aperture radar (SAR) imaging with millimeter-wave radar has been widely investigated in both academic research and industrial applications. However, traditional SAR imaging algorithms mostly focus on two-dimensional information of detected targets which restricts the development of 3D scenes reconstruction applications. In this paper, a novel low-cost 3D imaging algorithm based on 77 GHz commercial millimeter-wave radar is proposed by fusing the advantages of wideband radar, SAR, and multiple-input multi-output(MIMO) radar. Through processing real data collected by a 77 GHz MIMO-SAR, the proposed algorithm is verified successfully with extracted range-azimuth-height information and is capable of reconstructing the real-world 3D scenes with high spatial resolution.

The invention discloses a method for converting an L-shaped linear array into a planar virtual array. The method comprises the steps of building a space rectangular coordinate system xoy-z, enabling receiving antennas to be arranged along an x axis and a y axis, and forming the L-shaped linear array, constructing virtual array elements on the basis of the L-shaped linear array to form a virtual area array, and calculating the position and angle from the target to any array element of the virtual area array through geometric analysis by utilizing the position and angle relationship of the target on the array element of the L-shaped linear array, and completing the conversion from the L-shaped linear array to the planar virtual array, thereby realizing the same performance as a real array element. The virtual area array is generated on the basis of the L-shaped linear array, so that the number of array elements is reduced, the technical barrier that a traditional area array is realized by depending on a large number of entity array elements and complex hardware is broken through, and the problems that a planar array is complex in structure, high in power consumption, high in cost andthe like are fundamentally solved.

The invention discloses a microwave photon radar detection method and system based on light injection frequency modulation, and the method comprises the steps: enabling a single-frequency light signalwith the amplitude linearly changing along with time to serve as an injection light signal of light injection type laser, enabling the light injection type laser to work in a single-period oscillation state, and obtaining a linear swept-frequency light signal; converting the obtained linear swept-frequency optical signal into an electric signal, and transmitting the electric signal as a radar detection signal to a target; and taking the linear swept-frequency optical signal as a reference optical signal, performing optical domain frequency conversion processing on a target reflection electricsignal to obtain an intermediate frequency signal carrying target information, and extracting the target information from the intermediate frequency signal. According to the invention, broadband swept-frequency radar detection signal generation is realized through a light injection mode, signal parameters such as a radar working band and bandwidth can be rapidly and conveniently adjusted, and nonlinear distortion and signal-to-noise ratio loss introduced by traditional photon frequency mixing/frequency doubling are avoided; in addition, the system also has the advantages of simple structure and low cost.

The invention discloses a method of reducing the data transmission throughput of an FMCW (linear frequency modulated continuous wave) radar system. According to targets in the detection range of the radar system, the change distance of each target is limited within the preset small time window interval, the feature of spatial locality exists, each target corresponds to a distance dimension FFT, namely, 1st FFT, and the corresponding target frequency value on a frequency spectrum has invariance. According to the distribution condition of the concerned object targets in the 1st FFT, digital down-conversion processing of each frequency spectrum sub-band is carried out, namely, digital domain NCO shifting is carried out on the ADC (analog-to-digital converter) output, anti-aliasing filtering processing is carried out, then the corresponding multiple extraction output is carried out, and therefore the effect of reducing the data transmission throughput of the whole radar system is achieved.

The invention discloses a microwave photon time division multiplexing MIMO radar detection method and a system for realizing the detection method. The method comprises the steps that: a baseband linear frequency modulation signal is modulated to an optical carrier at a transmitting end, and a modulation optical signal comprising two sweep frequency components is generated through a photon frequency multiplication technology; the modulation optical signal is divided into two paths, one path of the modulation optical signal is sent to a radio frequency switch comprising M output ends after photoelectric conversion, each path of output of the radio frequency switch is connected with a transmitting antenna, and time division detection signal transmission of the M paths of array antennas is realized by controlling the on-off time sequence of the radio frequency switch; meanwhile, at a receiving end, the other path of the modulation optical signal is divided into N paths, and optical domaindown-conversion receiving is carried out on a target reflection signal; and M*N paths of intermediate frequency digital signals carrying target information are obtained in a work period, and the digital signals are processed to obtain detection target information. According to the invention, the azimuth angle resolution of a radar system can be improved while the high distance resolution is achieved.