Recognizing outdoor scene objects using texture features and probabilistic appearance model

Scene object recognition facilitates a large number of applications, ranging from indoor and outdoor, natural and man-made object recognition applications. In this paper, we propose a method for recognizing outdoor scene objects by using local features and contextual features. Local features consist of color, texture features extracted from objects in image then select which features best represent for each object using optimal feature subset selection algorithm. Objects features are modeled by a Gaussian distribution. Combining probabilistic spatial appearance of objects in images, probabilistic map from each N split-blocks of test image is generated. Blocks of high probability value are chose and using region growing method to segment the image. Objects can be recognized after fully segment the image. Effectiveness of the proposed method is verified through experiments.