基于Pos-FCN的红外图像背景抑制方法

红外图像背景抑制可以为红外目标检测识别任务提供支撑。在实际的应用场景中，红外图像中的目标多为弱小目标，其特征不明显，一般背景抑制算法难以将其从背景中分离，而达不到背景抑制的最佳效果。针对上述问题，提出使用Pos-FCN网络实现红外图像背景抑制的方法，该方法使用特征卷积结构，依靠高分辨网络结构获取弱小目标的特征信息，通过大尺寸卷积特征图的前向传播方式实现了高维度特征中弱小目标信息的保留，使用卷积降采样特征提取和上采样图像恢复方式实现了端到端的处理，并在前置训练阶段引入了位置信息强化网络骨干特征提取效果。结果表明，该方法处理后的红外图像中信杂比提高至3.877，对比度提高至0.297，检测率达到了93.6%，因此，该方法可以实现良好的背景抑制效果。

Background Suppression Method of Infrared Image Based on Pos-FCN

Infrared image background suppression can provide support for infrared target detection and recognition tasks. In practical application scenarios, the targets in infrared images are often small and weak, with unclear features. Generally, background suppression algorithms have difficulty separating them from the background, and thus cannot achieve the best background suppression effect. To address this issue, this paper proposes a method for infrared image background suppression using the Pos-FCN network. This method uses a feature convolution structure and relies on a high-resolution network structure to obtain feature information on small and weak targets. It realizes the preservation of small target information in high-dimensional features through the forward propagation of large-sized convolutional feature maps. It uses the convolutional downsampling feature extraction and upsampling image restoration to achieve end-to-end processing. In the pre-training stage, a position information-enhanced network backbone is introduced to enhance the feature extraction effect. The results show that the signal-to-noise ratio of the processed infrared image is increased to 3.877, the contrast is increased to 0.297, and the detection rate reaches 93.6%. Therefore, this method can achieve a good background suppression effect.