RGB-D Images for Objects Recognition using 3D Point Clouds and RANSAC Plane Fitting

in this paper, we highlighted object localization and recognition using RGB-D images that is top of RGB scenarios and provide semantically richer pixel-level support aps for individual object. Indeed, depth information levels with disparity-range of various objects in an image are used to extract objects of interest. Using proposed methodology, we extract point clouds from a depth image to proper plane fitting using Random Sample Consensus (RANSAC). RANSAC is challenging to handle the contour with thin edges. After local segmentation, we extracts various features like HOG and shape cues values to explore spatial properties of each object class. For object classification, we applied two well-known classifiers i.e., random forest (RF) and linear SVM. In the experimental evaluation, we achieved a gain of 16% relative improvement over current state-of-the-art methods. The proposed architecture can be used in autonomous cars, traffic monitoring and sports scenes.