This paper proposes an algorithm of affine-invariant starting-point matching of object contours,which leads to a new idea for fast object recognition.In the method,two affine-invariant arc-lengths based on curvature and enclosed area are employed to construct a characteristic function.Then,by utilizing extrema information of the characteristic function on amplitudes and positions,the amplitude error and relative-position error are derived based on which an estimation function is generated.According to cyclic-comparison and the rule of neighbouring position correspondence,the position variable which minimizes the estimation function is found,corresponding to the final resolved starting point of object.The experimental results show that the proposed method possesses high matching rate,high ability of anti-noise and low time-complexity.