Fig. 1. Refinement process of checkerboard corner subpixel refinement method based on edge direction projection. (a) Checkerboard image; (b) gray edge image; (c) intercepted neighborhood image; (d) initial edge orientation; (e) refinement of edge direction; (f) refinement of corner subpixels based on maximum projection of edge direction; (g) subpixel refinement of target corner; (h) complete refinement of all corner subpixels

Fig. 2. Pseudocode for obtaining local maximum of function

Fig. 3. Characteristics of corner neighborhood image and gray edge image. (a) Corner neighborhood image; (b) gray edge neighborhood image; (c) pixel screening results

Fig. 4. Schematic of binocular system and checkerboard edge length measurement. (a) Binocular system; (b) measurement requirements of checkerboard edge length (157 checkerboard edge length should be measured for each group of images, 80 in x direction and 77 in y direction)

Fig. 5. Test results of each method under high quality image. (a) Test results of B_max; (b) test results of B_mean

Fig. 6. Checkerboard image with noise disturbance. (a) Checkerboard pattern segment image under Gaussian noise; (b) checkerboard pattern segment image of corner contamination

Fig. 7. Test results of each method under Gaussian noise. (a) Test results of B_max; (b) test results of B_mean

Fig. 8. Test results of each method under corner contamination. (a) Test results of B_max; (b) test results of B_mean