The detection of the space-based infrared warning system usually dwells on the infrared images processing. A new detection method based on least squares support vector machine (LSSVM) combining spectral dimension information is discussed by analyzing the spectral dimension data and the low signal to noise ratio. Artificial bee colony algorithm is used to optimize the kernel function and regularization parameters, using F measure function as the fitness value. A LSSVM classifier is designed under small training set conditions, using the plume infrared spectrum of four typical missile types as training samples. The detection effect shows that the model of artificial bee colony algorithm optimization achieves better results compared with grid search algorithm, particle swarm optimization and genetic algorithm especially under low signal to noise ratio.