针对红外序列图像中弱小目标检测问题,提出了用最小二乘向量机对原始红外图像中每一像素的局部区域作灰度曲面最佳拟合,在拟合曲面上进行灰度极大值像素点位置估计,实现目标的粗定位。真正的目标取决于中心点的灰度是否高于其邻域的平均灰度。并以径向基核函数为例推导出了极值点估计所需的二阶方向导数算子。对模拟和实际图像进行了小目标检测的实验验证。结果表明,基于支持向量机的小目标检测算法具有较强的适应性。 In order to solve the problem of detecting weak targets in infrared sequence images, this paper proposes a method of least squares vector machine for the best fit of the local area of ​​each pixel in the original infrared image, and the gray-scale maximum pixel Point location estimation, to achieve the goal of coarse positioning. The real goal depends on whether the gray level of the center point is higher than the average gray level of its neighborhood. Taking the radial basis function as an example, the second-order directional derivative operator for extremal point estimation is deduced. Experimental verification of small target detection on simulated and real images. The results show that the small target detection algorithm based on support vector machine has strong adaptability.