HTHongxin Tan

Hongxin Tan

*This is an algorithmically generated author record based on publicly available information.

Hongxin Tan has a total of 13 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • SF-ISTD: A lightweight infrared small target detection model based on spatial-frequency domain feature enhancement

    by Lili Zhang, Ruiyang Xiao, Wenshuo Han, Yao Huang, Jing Li, Wei Wei, Hongxin Tan

    Infrared small target detection significantly improves the detection capabilities of unmanned aerial vehicles (UAVs) in complex environments. However, infrared small targets in complex environments often exhibit extremely blurred edges, leading to degraded detection accuracy. To address this challenge, this paper proposes SF-ISTD, a lightweight model that introduces a spatial-frequency domain feature enhancement framework for UAV-based infrared small target detection. It consists of three core modules: an edge extraction stem (EES), a lightweight reparameterized convolutional block (LRConv Block), and a spatial-frequency enhancement module (SFEM). EES extracts target edge information, thereby providing the network with effective and learnable features. The LRConv Block constructs a lightweight backbone network that preserves features of infrared small targets. SFEM enhances the discriminative ability for infrared small targets through the spatial-frequency domain. Experimental results on the high-altitude infrared thermal unmanned aerial vehicle dataset show that SF-ISTD achieves an mAP50 of 84.22% and an mAP50:95 of 54.15%, with a detection speed of 148 FPS. Furthermore, compared with the baseline, SF-ISTD reduces model parameters by 27.64% and computational cost by 32.10%. SF-ISTD achieves competitive performance compared with state-of-the-art methods.

  • Clairaudience: a lightweight attentional residual neural network with data augmentation and feature fusion for underwater acoustic target recognition

    by Jing Li, Yucheng Han, Xudong Wang, Xinglong Zhang, Lili Zhang, Wei Wei, Pei Yu, Hongxin Tan, Kang Yang

    Marine engineering has boomed and many deep learning-based methods have been proposed for underwater acoustic target recognition. However, most of these methods are dedicated to develop more complex convolutional neural networks to achieve better performance. This results in these models being unable to be deployed to low cost and miniaturized automatic underwater vehicles. A novel lightweight attentional residual neural network with data augmentation and feature fusion is proposed in this paper. Mel Frequency Cepstral Coefficient (MFCC), delta-MFCC and delta–delta MFCC features are extracted in the time dimension for fusion to obtain the fusion feature. The SpecAugment data augmentation is also used to enhance the randomness and diversity of features by masking in time and frequency dimensions randomly. Shuffle attention in the residual blocks is introduced to enhance the representation of features. The lightweight model is evaluated and compared by using several metrics on ShipsEar and DeepShip datasets. The proposed lightweight model only requires 1.628 M parameters for the trained model. This work shows that the proposed method requires small memory storage, while it achieved comparative performance.