Road analysis based on texture similarity evaluation

摘要: The paper presents an image processing algorithm based on statistical methods in order to evaluate the road delimiting by textured region similarity measurement and defect texture detection localization. With purpose of validation, images are divided sixteen equivalent regions. For proper identification classification, a decision theoretic method two types statistic feature used. first type features derive from medium co-occurrence matrices: contrast, energy, entropy, homogeneity, variance, but normalized form. second is edge density per unit area. algorithms implemented Visual C++ Matlab allows simultaneously display both investigated region, Euclidian distance between them reference region. basic (reference) considered asphalt one different textures like grass pebble. result classification tested non-road type, evaluation localization