Research On Automatic Measuring System Of Geometrical Parameters To The NC Tool Based On Vision

Nowadays the requirements to product precision become higher and higher in the processing and manufacturing industry,and the geometrical parameters of the tool are of significant influence to the machining precision.Currently the cutting tool detection equipments used in China have no abilities to reach the high accuracy processing standard.Owing to the above problems,machine vision detection technology was applied to the detection of NC tool,and the damages to tool are avoided for its non-contact detection mode,meanwhile,detection efficiency and accuracy are effectively improved.First,the mechanical structure utilized in cutting tool visual detection was developed,as the important premise to achieve high detection accuracy.Then,aimed at the adaptive detection of cutting tool with different dimensions and shapes,a teaching and reappearing detection strategy was designed and programmed.These theoretical analysis and experimental data established the foundation for later researches of visual detection of cutting tool geometrical parameters.Secondly,after analyzing the focusing methods,the autofocus method to get the images of cutting tool was achieved.Then,the Otsu threshold value was utilized to separate the cutting tool from background.Then tracking the image boundary,the boundary points of cutting tool were inspected,and a corresponding correction factor was proposed to achieve the edge location according to the differentiation between boundary points and edge points.This method still has a single response to the fuzzy cutting edge.Meanwhile,on the basis of gray gradient at the boundary point,the interpolation method had been used to realize the real-time detection of sub-pixel edge.And then,targeting at the deficiency of cosine curvature,one improved curvature algorithm was proposed,with this improved method,curving direction and curve rotation rate can be calculated at the same time.Then,deeply researching the contour curvature characteristics near different characteristic points,The application scope to the method had been determined according to the practical dimension of the tool,In order to achieve accurate positioning of characteristic points.At last,The feasibility and accuracy of this improved algorithm was verified via identification with contour characteristic points in ideal and actual images.Finally,considering that the positioning of contour characteristic points is difficult after tool was worn,this paper put forward a viewpoint that regards the feature points as the unknown quantity and then optimizes.Converts the segmentation fitting to nonlinear fitting,conducts space transformation to the contour so as to meet the fitting requirements,and constructs the fitting objective function with the piecewise function.Then,the trust-region method had been adopted for solution.Through selecting the initial value reasonably,the algorithm within the domain of definition could be converged to optimal solution,and also the nonlinear optimization method could retain the geometrical characteristics at the feature points by the time of reducing the fitting residual error had been proved.