A Computational Approach to Edge Detection
IEEE Transactions on Pattern Analysis and Machine Intelligence
| Hi-index | 0.00 |
Automated evaluation of surface roughness is essential in the current trends of industrial product quality monitoring systems. We have devised a method that can be implemented online to decide whether the surface finish has achieved the desired roughness value. We use discrete wavelet transform (DWT) and the dual-tree complex wavelet transform (DT-CWT) for analysing the surfaces manufactured by the machining processes namely Milling, Casting and Shaping. Important consideration here is the appropriate selection of features to characterize the surface texture. We propose combinations of texture descriptors namely standard deviation, kurtosis, the properties of gray-level co-occurrence matrix (GLCM) and the Canny edge descriptor to form a robust feature set. Further we use Canberra distance metric as a similarity measure. The algorithm and the results with both DWT and DT-CWT are presented. Two effective combinations of texture descriptors are found out. The feature set comprising standard deviation, kurtosis and GLCM gives the correct classification performance of 94.72% with DT-CWT and 79.66% with DWT whereas the feature set of standard deviation, kurtosis and Canny edge descriptor shows performance of 92% with DTCWT and 82.22% with DWT.