针对空中红外点目标的特性,提出了一种检测算法。首先,对图像进行高通滤波预处理,增强目标与背景的对比度。增强后的相邻帧图像做差分运算以进行目标的粗检测,提取出候选目标区域;其次,对原图像或预处理后的图像在候选区域内进行自适应分割,检测出点目标;最后,根据目标像素在序列图像中运动的连续性排除噪声干扰,从而准确地检测出真实目标。通过对红外图像序列进行实验,证实了算法的有效性。 Aiming at the characteristics of airborne infrared point targets, a detection algorithm is proposed. First, the image is high-pass filtered to enhance the contrast between the target and the background. Secondly, the original image or the preprocessed image is adaptively segmented in the candidate region to detect the point target. Finally, According to the continuity of the motion of the target pixel in the sequence image, the noise interference is eliminated, so that the true target can be accurately detected. Experiments on infrared image sequences confirm the validity of the algorithm.