2 May 2022 in Measurement
14 December 2023 in IEEE Sensors Journal
Onboard obstacle detection is a vital technology to ensure the safety of smart trains. Traditional object detection algorithms have poor detection accuracy for small obstacles and are challenging to cope with complex changes in rail transit environments. To address the above problems, a robust deep learning algorithm RFA-Net is proposed. The improved focal and global knowledge distillation is used to improve the feature extraction capability without additional computational burden. The side-aware boundary localization subnet is used to enhance the detection head to obtain high-quality detection bounding boxes. Experimental results based on a fully annotated rail transit dataset show that the RFA-Net achieves a detection accuracy of 92.7%mAP while keeping the model lightweight. Compared with the baseline model, the overall detection accuracy is improved by 5.8%, and the detection accuracy of small obstacles is improved by 7.5%. The RFA-Net has a TPR of 0.893 and a FPR of 0.016, achieving a good balance in identifying dangers and reducing false alarms. In various complex rail transit environments, the RFA-Net is able to detect obstacles stably, and the detection accuracy for small obstacles is improved by 18.6%. In addition, experimental results based on the KITTI dataset show that the proposed algorithm has good generalization ability in traffic scenes.
1 August 2013 in IET Radar, Sonar & Navigation
In recent years, cognitive radar (CR) with waveform diversity has exhibited significant performance improvements over the traditional fixed waveform radar and become an area of vigorous research and development. This study presents a dynamic waveform selection algorithm to strive for tracking error minimisation for CR manoeuvering target tracking in clutter. Based on the concepts of resolution cell and measurement extraction cell, the statistical characteristics of radar measurements are discussed without dependence upon the Cramér‐Rao lower bound of the measurement errors and the high signal‐to‐noise ratio assumption. A particle filter combined with probabilistic data association is used as a tracker. To quantify the utility of available waveforms, the predicted tracking mean‐square error, because of its dependence on actual future measurements, is approximated efficiently via Gaussian fitting of the prior density of the target state and statistical linearisation of the measurement equation. Monte Carlo simulation results show that the proposed dynamic waveform selection algorithm can improve tracking performance considerably, especially in terms of track loss probability.
6 October 2016 in IEEE Antennas and Wireless Propagation Letters
To deal with the perfect electric conductor rough targets’ scattering problem, which is the basic premise of the radar applications at terahertz frequencies, a hierarchical semideterministic modeling and full-wave-algorithm-based computation method was proposed. The scattering fields with phase information and synthetic aperture images of electrically large cubes with different surface characteristics were obtained. The influence of the surface roughness can be easily observed. According to the imaging results, we can predict that other than the scattering calculation challenges caused by the surface roughness, it also brings great benefits that more detailed information about the targets will emerge on the images.