Provided are a system and method for real-time object recognition and pose estimation using in-situ monitoring. The method includes the steps of: a) receiving 2D and 3D image information, extracting evidences from the received 2D and 3D image information, recognizing an object by comparing the evidences with model, and expressing locations and poses by probabilistic particles; b) probabilistically fusing various locations and poses and finally determining a location and a pose by filtering inaccurate information; c) generating ROI by receiving 2D and 3D image information and the location and pose from the step b) and collecting and calculating environmental information; d) selecting an evidence or a set of evidences probabilistically by receiving the information from the step c) and proposing a cognitive action of a robot for collecting additional evidence; and e) repeating the steps a) and b) and the steps c) and d) in parallel until a result of object recognition and pose estimation is probabilistically satisfied.