Supervised texture classification using several features extraction techniques based on ANN and SVM

Texture-based Classification of Workpiece Surface Images using the Support Vector Machine

“…A Gabor filter is composed of a sinusoidal plane wave of a particular frequency and orientation, modulated with a Gaussian kernel [3]. The mathematical representation can be written as: …”

“…The SVM was trained using Platt"s Sequential Minimal Optimization (SMO) algorithm. SMO was chosen due as it proved to be effective in our previous work [3].…”

“…Texture differences are perceived in object structure, roughness, orientation, smoothness and regularity, which are also commonly unique across processed workpieces [1]. The uses of texture analyses are widespread, such as for classification [2,3], retrieval [4], and inspection [5].…”

“…The same feature was also used in [24] for establishing the relationship between milled material roughness and its underlying surface texture. In [3], machining processes were categorized via analysis of gray-level histogram information, combined with edge statistics from metal workpiece images.…”

“…Among the algorithms that have been used are decision trees [11], Artificial Neural Networks (ANN) [3,7,10] and the Support Vector Machine (SVM) [3,12,17].…”