Satellite cloud image processing is widely used in meteorology, and convective cloud attracts great attentions in me-teorological monitoring. Generally speaking, convective cloud plays a pivotal role in governing the rainfall, and they are also responsible for modulating the radiation budget of earth atmosphere system. Especially, the emergence of cumulonimbus which generates at the beginning of convection is often indicating thunder and lightning, torrential rains or even accompanies typhoons and other natural disasters. Hence, the convective clouds detection is a key factor for weather forecasting, climate monitoring and helps to prevent natural disasters.In this paper, a modified Support vector machine (SVM) was proposed to detect convective clouds. The traditional SVM is easily affected by noises and outliers, and its training time will dramatically increase with the growing in number of training samples. On the other hand, satellite cloud image may easily be deteriorated by noises and inten-sity non-uniformity with a huge amount of data needs to be processed regularly, so it is hard to detect convective clouds in satellite image using traditional SVM. To deal with this problem, a novel method for detection of convec-tive clouds based on a fast fuzzy support vector machine (FFSVM) was proposed. FFSVM was constructed by elim-inating feeble samples and designing new membership function as two aspects. First, according to the distribution characteristics of fuzzy inseparable sample-set and the fact that the classification hyper-plane is only determined by support vectors, this paper uses SVDD, Gaussian model and border vector extraction model comprehensively to de-sign a sample selection method in three steps, which can eliminate most of redundant samples and keep possible support vectors. Then, by defining adaptive parameters related to attenuation rate and critical membership on the basis of the distribution characteristics of training set, an adaptive membership function is designed. Finally, the FFSVM was trained by the remaining samples using adaptive membership function to detect convective clouds. The experiments on FY-2D satellite images show that the proposed method, compared with traditional FSVM where no samples were eliminated, not only remarkably reduces training time, but also further improves the accuracy of con-vective clouds detection.