To measure the rounded corner and diameter of a micro-drill, the fitting of the elliptical equation for margin projection of the micro-drill was researched.Firstly,the margin projective image was collected with a high precise automatic test system of PCB micro-drill. While the elliptical equation of margin projection of micro-drill was fitted, the sum of square distance from the sample points’ vector coordinates to the hyperplane determined by the elliptical coefficients was taken as an objective function. Then, a Sanger neural network with lateral connection was designed, in which the input signal were vector coordinates composed of sample points.Finally,on the basis of an adaptive minor component extracting method,the eigenvector of minimum eigen values was taken as the fitted coefficients of the hyperplane, i.e., fitted elliptical coefficients, and the rounded corner and diameter of the micro-drill were obtained. The proposed approach can meet the precision requirements of micro-drill testing, and is a novel application of adaptive Sanger network to measuring shape errors.