Abstract: Taking the conical thread as the object of study, a subpixel image detection method for the mechanical part linear edge based on the support vector regression was proposed. Through processing the conical thread image recorded by the charge coupled device(CCD), the pixel characterization of the straight portion of the thread form could be obtained with the algorithm of median filter, binary transform etc, to make up the training set. The support vector machine for regression was trained by the training set and the straight line equation of the thread form was obtained which is the expression of the subpixel . Making use of it to detect the main parameters of the conical thread, the effect of the discreteness of the CCD and the system noise on the detection results were significantly reduced. The experiment results show that the suggested method is characterized by high detection speed and high precision.

Key words: computer application, image detection, CCD, subpixel, support vector regression