Fig. 1. Schematic of BRDF simulation

Fig. 2. Imaging path of object in near-field photometric stereo vision scene^[8]

Fig. 3. Quasi point source emitting light in any direction

Fig. 4. Metal surface texture reconstruction process based on near-plane near-field photometric stereoscopic vision

Fig. 5. Diagram of shooting scene

Fig. 6. Near-field non-Lambert photometric stereo vision network based on co-location light source

Fig. 7. Schematic of near-field photometric stereo vision device

Fig. 8. Simulation results. (a) True value of normal vector; (b) predicted value of normal vector; (c) angle error of normal vector; (d) truth value of surface texture; (e) surface texture reconstruction value; (f) surface texture error; (g) sectional profile extracted from Fig. (d); (h) sectional profile extracted from Fig. (e)

Fig. 9. Error statistics of simulation results. (a) MAE of surface normal vector; (b) RMS relative error of surface texture depth value

Fig. 10. Experimental results of stainless steel end milling sample No. 1. (a) Sample photographs; (b) normal vector graph; (c) surface textures as measured by white light interferometer; (d) surface texture measured by proposed method; (e) sectional profile from Fig. (c); (f) sectional profile from Fig. (d)

Fig. 11. Experimental results of aluminum alloy end milling sample No. 2. (a) Sample photographs; (b) normal vector graph; (c) surface textures as measured by white light interferometer; (d) surface texture measured by proposed method; (e) sectional profile from Fig. (c); (f) sectional profile from Fig. (d)

Fig. 12. Experimental results in different areas of field of view. (a) Measurement of regional distribution; (b) surface textures as measured by white light interferometer; surface textures measured by proposed method in (c) region 1, (d) region 2, (e) region 3, (f) region 4, and (g) region 5

Table 1. Statistics of different measurement results

Table 2. Statistics of measurement results in different areas