Classification for Striation Patterns Using the Synthetical Feature Vector Based on SVM

With the advent of high-efficiency technology of digital image processing and pattern classification, the research on classification for tool marks is catching forensic scientist's eyes. It is crucial for classification to extract and select the features from tool marks. In the practical situation, the geometrical shapes and the textures of tool marks are complex, irregular and stochastic. It is difficult to represent the tool mark using a single feature. A new approach of the feature extracting and representation is presented. It computes multi-scale extended fractal features and morphological structure features which are constructed into a synthetical feature vector. The vector is utilized to classify the striation patterns after the reduction of its dimensionality based on SVM. Experimental result shows that the method presented is effective for classification of striation patterns.