The detection and evaluation of the corrosion characteristic parameters of steel bars is the basis for the performance evaluation of corroded reinforced concrete structures. The corroded steel bars with different target corrosion degrees were obtained by electrochemical accelerated corrosion test. Three-dimensional laser scanning technology was adapted to establish 3D geometric models of corroded steel bars, and the relevant corrosion characteristic parameters were extracted. Then the relationship between the average and maximum cross-sectional area loss ratio, as well as the corrosion non-uniformity coefficient were analyzed based on the probability theory. The probability distribution model of corrosion non-uniformity coefficient was also established. The results show that with the increase of average cross-sectional area loss ratio, both the maximum cross-sectional area loss ratio and the corrosion non-uniform coefficient increase as power function, and the parameters of power function are dependent on the diameter of the corroded steel bar. The corrosion non-uniform coefficient could be characterized by the generalized extreme value distribution. The shape, scale, and location parameters of the probability distribution increase linearly with the increase of corrosion degree, which are also dependent on the element length and the diameter of the corroded steel bar.