Because of the influence of nature meteorological condition, background environment and the structure of target, the detection of weak and small targets in infrared image is one of the difficulties among image targets detection and identification.A new small target detection method is presented,in which the infrared image intensity surface is well fitted by the least square support vector machine （LS- SVM）, and then the maximum extremum points are detected based on the fitted intensity surface. It is decided whether the extremum point is a target position or not by comparing the gray level of the point with the average gray intensity of its local area. The second order directional derivative operators are deduced from the LS-SVM with the radial basis kernel function, as an example. The computer small target detection experiments are carded out for the real and simulated images. Experimental results demonstrate that the proposed algorithm is robust and efficient.