Abstract: Sizing of cracks is one of the difficult points of the quantitative nondestructive testing. For a Stress Corrosion Crack(SCC), the reconstructed crack profile is often underestimated especially for the parameter of crack depth because of the ill description of the corresponding inverse problem. In this paper, a fine crack model, which consists of a group of crack zones of different conductivity value, is proposed for the numerical simulation of SCC signals based on the measured conductivity distribution of a SCC. The reconstruction precision can be effectively improved by utilizing the new crack model and a hybrid inversion scheme of the Particle Swarm Optimization(PSO) algorithm and the Conjugate Gradient(CG) method. Reconstruction results from simulated and the measured eddy current signals of an artificial SCC demonstrate that the proposed numerical model and the inverse analysis scheme are effective for the sizing of SCC.