A detection method of small target in infrared image is proposed, which is based on the background predication by fuzzy least squares support vector machine (FLS-SVM) and threshold segmentation by fuzzy Tsallis-Havrda-Charvat entropy. Firstly, the fitting function is obtained from the training samples by using FLS-SVM and the background in infrared image is predicted. Then, the predicted image subtracted from the source image gives the residual-error image. The residual-error image is segmented by the proposed threshold selection method based on fuzzy Tsallis-Havrda-Charvat entropy so as to separate small target and noise from the residual background. Finally, the true small target is further detected based on the stability of the target gray and the consistency of target trajectory. The experimental results are given and analyzed. They are compared with the detection results of the background predication methods based on LS-SVM or least squares. The results show that the proposed method has higher detection probability and the gain of signal-to-noise ratio (GSNR) and it is superior to the above-mentioned methods.