Subspace quadratic synthetic discriminant function(SSQSDF) is one type of quadratic correlation filters which show well performance for target detection.In order to gain better results on IR target detection under complex scenery,a kernel SSQSDF based algorithm(KSSQSDF) is proposed,which is an extension of SSQSDF from low dimensional space to high space is proposed.The kernel matrix of training samples is first obtained with suitable kernel function,then SSQSDF is transformed to kernel martrix equation using property of kernel function,at last the expression of KSSQSDF is achieved.The proposed algorithm can highly restrain background noise,and improve detection precision.The experiments on real scene verify that KSSQSDF outperforms SSQSDF in detecting infrared targets.