Intelligent Vision Based Technique Using ANN for Surface Finish Assessment of Machined Components
In this work, an FPGA hardware based image processing algorithm for preprocessing the images and enhance the image quality has been developed. The captured images were processed using a FPGA chip to remove the noise and then using a neural network, the surface roughness of machined parts produced by the grinding process was estimated. To ensure the effectiveness of this approach the roughness values quantified using these image vision techniques were then compared with widely accepted standard mechanical stylus instrument values. Quantification of digital images for surface roughness was performed by extracting key image features using Fourier transform and the standard deviation of gray level intensity values. A VLSI chip belonging to the Xilinx family Spartan-IIE FPGA board was used for the hardware based filter implementation. The coding was done using the popular VHDL language with the algorithms developed so as to exploit the implicit parallel processing capability of the chip. Thus, in this work an exhaustive analysis was done with comparison studies wherever required to make sure that the present approach of estimating surface finish based on the computer vision processing of image is more accurate and could be implemented in real time on a chip.
Guo Fan JIN, Wing Bun LEE, Chi Fai CHEUNG and Suet TO
M.R. Narayanan and S. Gowri, "Intelligent Vision Based Technique Using ANN for Surface Finish Assessment of Machined Components", Key Engineering Materials, Vols. 364-366, pp. 1251-1256, 2008