Sub-pixel straight lines detection for measuring through machine vision

Abstract

External visual interfaces for high precision measuring devices are based on the segmentation of images of their measuring reticle. In this paper, a method for subpixel straight lines detection is presented and tested on images taken from the reticle of a dark field autocollimator. The method has three steps, the sharpening of the image using a version of the Savitzky-Golay filter for smoothing and differentiation, the construction of a coarse edge image using Sobel filters, and finally, the subpixel edge location determination, by fitting a Gaussian function to orthogonal sections of the coarse edge image.We discuss results of applying the proposed method to images of the reticle of a Nikon 6D autocollimator, using the scale of the device as a benchmark for testing the error in the location of the lines and compare them with Sobel/Hough and Sobel/polynomial fitting. We report that for this type of image-content, Gaussian fitting has smaller uncertainty, when cameras with two different sensors are used.