Water‐surface infrared small object detection based on spatial feature weighting and class balancing method

• Published in: IET image processing 2023-08, Vol.17 (10), p.3012-3027
• Main Authors: Hui, Tian, Xu, YueLei, Li, HuaFeng, Zhou, Qing, Rasol, Jarhinbek
• Format: Article
• Language: English
• Subjects: computer vision; convolutional neural nets; feature extraction; Gaussian distribution; image processing; infrared detectors; object detection
• ISSN: 1751-9659; 1751-9667
• DOI: 10.1049/ipr2.12851

Infrared imaging is widely used due to its penetration capability to operate under many weather or lighting condition. However, due to the far distance of aerial view, feature blur, and the scarcity of aerial infrared data, the detection of small infrared targets on the water surface remains a challenging problem. In response to the problem of unclear features, we propose the spatial feature weighting method based on 2D Gaussian distribution. This method increases the weight of the target area by adaptively adjusting the feature activation. Secondly, for the problem of rare aerial perspective infrared data, we propose the cross‐spectral data migration method. By introducing the domain difference loss function to optimize the pseudo‐label selection process, the range of target domain distribution is expanded, and the adaptability of the detector is improved. Finally, in response to the problem of underfitting caused by category imbalance in transfer learning, we propose the class balancing method that effectively reduces the false detection. Extensive experiments were conducted on both benchmark datasets and the self‐built dataset to evaluate the effectiveness and robustness of our method. The proposed method was evaluated with different models and various scenarios, and the results demonstrated the effectiveness. We tackle challenges posed by unclear features of small targets, sparse data, and high false detection rates. We introduce the spatial feature weighting which utilizes 2D Gaussian distribution to assign adaptive weights to regions. We present cross‐spectral data migration method that optimizes the process of pseudo‐labels. We propose the class balancing method to mitigate the high false detection rate.