In order to realize the high-accurate vision measurement of large scale machine parts

the problem that at present

the range of vision measurement is too small must be solved and the effects of image distortion on measurement accuracy must be managed to get rid of; Accordingly

a vision measurement method based on dimension characteristics of sequential partial images is proposed. Firstly

the transition distribution modality of part image edges and its effects on measurement accuracy are analyzed. To get rid of the effects

a method of edge pixel compensation is proposed

which availably improve measurement accuracy. Then

taken the dimension measurement of straight edge parts as prototype

the vision measurement method based on dimension characteristics of sequential partial images is introduced. This method consists of three steps. First

take the images of small areas in the part and produce the sequential partial images which overlap at adjacent margins one by one. Second

apply the cross-correlation matching technique and the bilinear interpolation algorithm to obtain the dimension characteristic lines of the adjoining sequential images in sub-pixel level. On the basis of it

extract the dimension characteristic of each partial image and perform sum and compensation of these sizes to get the dimension of the part. The experiment shows that

applying the edge pixel compensation method to the single image of general scale parts

the relative error is within 0.008%; applying the sequential image measurement method to large scale parts

the relative error is within 0.01%. These indicate that the edge pixel compensation method can effectively get rid of the effects of unbeknown exact location of the real edge on measurement accuracy; by limiting the image area in the center position of the photosurface

the sequential image measurement method can be used to measure large scale parts with small distortion images. It has the advantage of small error accumulation. The measurement accuracy meets the need of the accurate measurement. Edge pixel compensation method and sequential image measurement method novelly promote the precision of machine vision measurement and can be applied to the precise auto-measurement of sheet metal parts. The thought of “dividing the whole into pieces and sum the pieces making the whole” is of great value for extending the application of machine vision measurement.