This paper applies least square-support vector machine(LS-SVM) to quantify the defects on tank floor of oil and gas, this method build the relationship between the three-axial magnetic flux leakage(MFL) of defects and the length, width and depth of defects. In order to accurately quantify the defect, particle swarm optimization（PSO）is adopted to optimize the model parameter of LS-SVM. According to the simulation results, PSO-LS-SVM needs less training time and has better accuracy for the quantification of defects than BP neural network method, and PSO-LS-SVM has better application advantages on the engineering.