As for the problem of small target detection in hyperspectral remote sensing image, a detection method based on adaptive parameter support vector machine (SVM) is proposed. The low dimensional information of the hyperspectral image is obtained using the method of principal component analysis (PCA) and the redundancy of data is reduced. Then, small targets are positioned fast and roughly by an unsupervised detection method, and the posterior information of SVM is got by the position result. The kernel parameter of SVM is determined adaptively based on the posterior information and the criteria of divergence in the kernel space. The best hyperplane in the kernel space for the segmentation of targets and background is found by the SVM. Pixels are separated to targets and background by the best hyperplane. The accurate and stable target detection result is obtained by iteration. A large number of experimental results show that, compared to the existing methods such as RX method, kernel RX method and support vector data description (SVDD) method, the proposed method is more effective to detect small targets accurately in the hyperspectral remote sensing image.