First, we regularize the attraction force field in the geometric active contour model to extend the capture range of the object boundaries, and improve the ability of convergence to the concavities.

A theoretical consideration to estimate the convergence conditions in learning control process is proposed in accordance with the limit condition of a geometric series.