针对复杂背景下红外图像中低信噪比弱小目标实时检测问题，提出一种基于相关滤波器的红外弱小目标检测算法。该算法将红外目标检测转化为模式分类问题，在离线训练阶段，利用二维高斯模型构造红外小目标训练集，在此基础上训练得到对目标背景具有区分能力的相关滤波器，在线检测阶段，利用滤波器对图像分块进行滤波操作，目标和背景的滤波响应有着显著的差异，最后生成整幅图像的滤波响应置信图以此来判断图像中是否包含目标及其具体位置。在单帧单目标图像、序列图像多目标检测实验结果表明，与经典检测算法相比，所提方法不仅具有更高检测性能，有效降低了虚警概率，而且具有较好的实时性，适用于复杂背景条件下弱小目标的实时检测。

To solve the infrared target real-time detection problems caused by low signal to noise ratio with complex background, an infrared small target detection method based on correlation filter is proposed. The small target detection problem is transformed into pattern classification task, which consists of two stages: off-line training and on-line detection. In training stage, a correlation filter is obtained using the training dataset produced by two-dimensional Gaussian model. It has ability to distinguish target and background. In detection stage, the sub-image blocks of the infrared image are extracted successively and the filtering response confidence map which indicates the target location is computed. The experiments under two conditions demonstrate that the proposed method not only has better detection performance with effective reduction of the false alarm rate, but also has better real-time performance.