The continuous advancement of shipbuilding technology has enabled the further development of sea transport and naval vessels, which also poses significant challenges to navigation safety. Accurate recognition of different types of ships is crucial. In this study, three typical ships, namely a destroyer, cruise ship, and aircraft carrier, were taken as examples. The micro-Doppler effect of these ships was analyzed using a unmanned aerial vehicle (UAV)-based ultrawideband (UWB) radar. Echo signal models were established and compared with measured Doppler signals. Furthermore, a lightweight self-supervised learning network was proposed to achieve accurate recognition of small-sample labeled ship targets. The neural network architecture, including specific layers and parameters, was introduced. To enhance the micro-Doppler features, the 1-D signals were transformed into 2-D images, thereby converting the ship recognition problem into an image classification problem. The experimental results demonstrate that the proposed network achieves high recognition accuracy under different situations. When evaluated on measured data, the overall recognition rate reaches 97.50%. Moreover, even in challenging scenarios with a low signal-to-clutter ratio (SCR) of −10 dB, the overall recognition rate for simulated data still exceeds 90%, demonstrating the effectiveness and advantages of the proposed network.

The recognition of ice channels plays a crucial role in developing intelligent ship navigation systems in ice-covered waters. Navigating through ice channels with the assistance of icebreakers is a common operation for merchant ships. Maneuvering within such narrow channels presents a significant challenge for the captain’s skills and ship performance. Therefore, it becomes essential to explore methods for enabling ships to navigate through these channels automatically. A key step in achieving this is the accurate recognition and extraction of boundary lines on both sides of the ice channel. An ice channel line recognition method based on the lane line detection algorithm UFAST is implemented. The method is trained and tested on the constructed ice channel dataset, with the test results showing that the average recognition accuracy reaches 84.1% and the recognition speed reaches 138.3 frames per second, meeting the real-time requirement. In order to solve the current lack of authentic ice channel images, ice channel navigation scenes are built based on UE4, and synthetic ice channel images are rendered. The method in this paper is also compared with the traditional non-intelligent Otsu threshold segmentation method and the intelligent instance segmentation method YOLACT for performance analysis. The method in this paper has 9.5% higher ice channel recognition accuracy and 103.7 frames per second higher recognition speed compared with YOLACT. Furthermore, ablation studies are conducted to analyze the relationship between the number of gridding cells in the proposed method and ice channel recognition accuracy.

Linear frequency modulation (LFM) radar which can measure distance of target is the key sensor in autonomous driving system in the context of Internet of Things (IoT). However, huge amount of raindrops in rainfall environment would reflect signal and result in negative influence. In order to improve the detection ability of LFM radar, echo signal model of LFM radar target in rainfall environment is established. Working principle of LFM radar, characteristics of raindrops, influence of rain clutter and comparison with measured data in time-domain and frequency-domain distribution are given in sequence. Secondly, three features are defined and extracted from frequency-domain spectrum and singular-domain spectrum to recognize target echo signal, rain clutter and mixed signal with high precision. When the rainfall rate is 3.33 mm/h, the overall recognition accuracy can still be up to 95.15%. Thirdly, an adaptive ensemble empirical mode decomposition (EEMD) algorithm is proposed to suppress rain clutter in mixed signal. Simulation results prove the effectiveness of EEMD in improving ranging accuracy in rainfall environment. At last, outfield experiments in normal environment and rainfall environment are conducted respectively. Measured distance obtained by proposed method in rainfall environment is still more accurate than that obtained by traditional fast Fourier transform (FFT) method in normal environment, illustrating that the proposed method can ensure the high performance of LFM radar in rainfall environment. Research in this paper has positive influence on improving reliability of LFM radar in inclement environment and can contribute to more stable target information for the whole in-vehicle IoT.

With the rapid advancement of neural networks and radar sensing, millimeter-wave radar-based human activity recognition (HAR) has become pivotal in sensing applications including robotics, edge-enabled smart homes, and autonomous driving. However, existing methods suffer from high computational costs and long training time, hindering their deployment on resource-constrained embedded systems. To address this, we propose a fast and robust HAR framework that integrates multidimensional micro-Doppler feature fusion with an edge-optimized Broad Learning System (BLS). Firstly, human activity data are collected using a millimeter-wave radar, and spectral features are extracted through a two-dimensional Fast Fourier Transform (2D-FFT). Subsequently, for the spatial-temporal characteristics of HAR, the pre-activation residual module, multi-scale convolution module, convolutional block attention module and SoftPool are utilized to design a lightweight Hybrid-CNN architecture. Finally, BLS is used for rapid learning and classification, constructing an efficient recognition model. Experimental validation demonstrates three key advantages: Recognition accuracy of 99.02% for nine poses by spectral-convolutional feature fusion; Noise robustness achieving 88.05% accuracy at -5dB SNR via adaptive SoftPool mechanisms; Data efficiency attaining 95.76% accuracy using only 660 training samples through BLS incremental learning. Cross-platform deployment validation demonstrates the method’s efficient execution capabilities on wearable devices and embedded computing units, with architectural optimizations significantly reducing resource consumption and meet real-time deployment requirements. The proposed method provides a sensor-adaptive solution for realizing fast and accurate human pose recognition in edge computing scenarios with limited computational resources.

Real-time angle measurement of missiles is as important as range measurement for precision strike in modern wars and fuse using linear frequency modulation (LFM) system has been equipped widely due to its good range resolution. However, angle measurement of fuse using LFM system hasn’t been realized. In order to achieve the accurate and rapid measurement of angle, this paper firstly derives the relationship between angle and phase of beat signal in the LFM system. Secondly, an arithmetic called AapFFT-CZT is proposed by sampling and inserting zeros based on all-phase Fourier Transform (apFFT) and Chirp Z Transform (CZT). Compared with traditional Fast Fourier Transform (FFT), AapFFT-CZT has higher accuracy and better anti-noise performance when used to measure angle. Thirdly, angle measurement utilizing AapFFT-CZT is performed successfully on single core of TMS320C6678. When there is no noise, accuracy of angle measurement is within 0.01°. When there is Gaussian white noise, accuracy is still within 0.25°. Last but not least, fast and parallel measurement on multiple cores is achieved. The results show that clocks consumed when using multiple cores is just 57% of clocks consumed when using single core. Therefore, angle can be measured accurately and quickly based on C6678 multi-core DSPs.