On the basic of computer vision principle

a surface measurement system of sheet metal parts is proposed in this paper. Using neural network

the mapping relation between image points and special points is established. Some distorted stripes are obtained on surface

and the points of stripe edges are detected by wavelet edge detection. A searching non supervisor clustering algorithm is discussed

so all of edge points are divided into different groups according to stripe edge situation

the edge points of every group are fitted into a B spline curve. The curves are recognized and marked based on geometric invariance to search corresponding points at sub pixel level. Furthermore

the multi scale and multi resolution attributes of wavelet are applied to image mosaic and data integration

so a large scale surface can be measured. At last the coordinates of points on surface are calculated with the mapping relation between image points and special points

and the measuring precision is less than 0 5mm/m.The system avoid emending optical system distortion of cameras

achieve stereo matching at sub|pixel level

and integrate surface data

so the large surface can be able to measure.