Abstract
The data-driven method for infrared small target detection (IRSTD) hasachieved promising results. However, due to the small scale of infrared smalltarget datasets and the limited number of pixels occupied by the targetsthemselves, it is a challenging task for deep learning methods to directlylearn from these samples. Utilizing human expert knowledge to assist deeplearning methods in better learning is worthy of exploration. To effectivelyguide the model to focus on targets' spatial features, this paper proposes theLocal Contrast Attention Enhanced infrared small target detection Network(LCAE-Net), combining prior knowledge with data-driven deep learning methods.LCAE-Net is a U-shaped neural network model which consists of two developedmodules: a Local Contrast Enhancement (LCE) module and a Channel AttentionEnhancement (CAE) module. The LCE module takes advantages of prior knowledge,leveraging handcrafted convolution operator to acquire Local Contrast Attention(LCA), which could realize background suppression while enhance the potentialtarget region, thus guiding the neural network to pay more attention topotential infrared small targets' location information. To effectively utilizethe response information throughout downsampling progresses, the CAE module isproposed to achieve the information fusion among feature maps' differentchannels. Experimental results indicate that our LCAE-Net outperforms existingstate-of-the-art methods on the three public datasets NUDT-SIRST, NUAA-SIRST,and IRSTD-1K, and its detection speed could reach up to 70 fps. Meanwhile, ourmodel has a parameter count and Floating-Point Operations (FLOPs) of 1.945M and4.862G respectively, which is suitable for deployment on edge devices.