In order to meet the fast and high-precision recognition requirements of feature points of positioning tool in CCD camera projection measurement

an improved k-cosine curvature recognition algorithm is proposed. An angle curvature calculation method by fixing support field is used in the algorithm to obtain the rotation rate and the bending direction of the curve. Firstly

the tool contour feature points are classified into LL-C type

AL-T type and AL-C type. Then

the curvature characteristics around different feature points are analyzed to achieve the accurate positioning of feature points. Finally

in order to verify the effectiveness of the algorithm

simulation experiments are conducted by standard contour model

the results show that the recognition and positioning errors of the method on tool feature point are less than 2 pixels.