Object Curvature Measurement Based On Machine Vision

China is a large country in industrial production and manufacturing.In order to produce necessary industrial components,the inspection of industrial components is essential in the process from production to sales.The qualities of components meet the quality standards determines whether they are recycled or sold.With the growth of advanced manufacturing industry,industrial output is increasing day by day,which indicates the increasing demand for product inspection.The conventional manual inspection cannot adapt this situation,since it.has the shortcomings of hard recognition only via human eyes,low efficiency,high labor costs,large subjectivity,and inability to work continuously with high intensity.In recent years,with the rapid growth of computer vision,machine vision has been successfully applied to the field of industrial inspection.In order to overcome the disadvantages of traditional manual inspection and the limitations of existing methods in efficiency and applicability,this thesis proposes a multibeam optics method for measuring the curvature of the object via machine vision,which is based on the basic principle of photometric stereo vision technology.Specifically,it involves a series of related image processing algorithms,such as image acquisition,image preprocessing,surface normal vector calculation,surface depth information calculation,surface restoration and inspection.First,an image acquisition platform is established for the curvature measurement via multi-beam optics.The images are acquired by the image acquisition system from multiple shooting angles.Then,denoising is performed on the images to suppress unnecessary noise interferences,and estimate the direction of the rotating light source for the denoised image and the distribution change of the corresponding feature point group of each image.Next,according to the relevant optical theories,the surface normal vector is achieved based on the relationship of the light source direction,the surface normal vector and the image pixel intensity.Due to the beam splitting of reflected light during data acquisition,the data is discrete and local,so the normal vector of the object surface is also discrete and local.Therefore,the solved surface normal vectors are further separated and interpolated.Finally,according to the relationship between the surface normal vector and the vertical section of the object,the depth information of the object surface is solved to restore the object curvature surface,which can be employed for curvature measurement.The experimental results show that the proposed surface curvature inspection method can meet the demand of industrial inspection,which achieves good performance in accuracy and inspection time.Specifically,the proposed method can measure the radius of curvature in 99.11% accuracy at the speed of 10.266 s,which provides an alternative method for surface curvature inspection of the industrial components.