A Dual Domain Approach for Surface Roughness Evaluation

The research presented in this paper is aimed at developing an automated imaging system for classification of engineered surfaces with appropriate roughness measures. The method uses dual domain of Radon transform and wavelet transform. Radon transform helps to locate directional information of surface textures whereas wavelet transform is used to compute features from textures. To extract significant information from textures we first employ Canny edge detection to texture image and then undergo Radon and wavelet transformation. Further we carryout the process on a set of four images original and rotated to form the robust feature set. Our feature vector has 36 features. Experimentation is carried out on three surface texture databases manufactured by machining processes milling, casting and shaping. We achieved 95.56% correct classification performance for milling samples where as 67.04% for casting and 87.08% for shaping samples.