SYSTEM AND METHOD FOR REAL-TIME OBJECT RECOGNITION AND POSE ESTIMATION USING IN-SITU MONITORING
First Claim
1. A method for real time object recognition and pose estimation using in-situ monitoring, comprising the steps of:
- a) receiving 2D and 3D real time image information, extracting single or multiple evidences from the received 2D and 3D real time image information, recognizing an object by comparing the extracted evidences with model information, and expressing locations and poses of the object by probabilistic particles in a space;
b) probabilistically fusing various locations and poses of the object, which are generated in a particle form, and finally determining a location and a pose of the object by filtering inaccurate information;
c) generating a region of interest (ROI) by receiving 2D and 3D real time image information and the location and pose of the object from the step b) and collecting and calculating real time environmental information;
d) selecting an evidence or a set of evidences probabilistically by receiving information from the step c) and proposing a cognitive action of a robot for collecting additional evidence if selected evidence is not sufficient; 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.
1 Assignment
0 Petitions
Accused Products
Abstract
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.
-
Citations
22 Claims
-
1. A method for real time object recognition and pose estimation using in-situ monitoring, comprising the steps of:
-
a) receiving 2D and 3D real time image information, extracting single or multiple evidences from the received 2D and 3D real time image information, recognizing an object by comparing the extracted evidences with model information, and expressing locations and poses of the object by probabilistic particles in a space; b) probabilistically fusing various locations and poses of the object, which are generated in a particle form, and finally determining a location and a pose of the object by filtering inaccurate information; c) generating a region of interest (ROI) by receiving 2D and 3D real time image information and the location and pose of the object from the step b) and collecting and calculating real time environmental information; d) selecting an evidence or a set of evidences probabilistically by receiving information from the step c) and proposing a cognitive action of a robot for collecting additional evidence if selected evidence is not sufficient; 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. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21)
-
-
22. A system for real time object recognition and pose estimation using in-situ monitoring, comprising:
-
an image capturing unit for continuously capturing an object to recognize with a real peripheral environment in various view points; a real time environment monitoring unit for receiving 2D and 3D image information from the image capturing unit and calculating and collecting real time environmental information; a multiple evidence extracting unit for extracting various evidences from the real time image information and generating various locations and poses of the object by comparing the extracted evidences with model information; an evidence selecting and collecting unit for selecting a best evidence for the object and the peripheral environment thereof using the real time environmental information and the model information and proposing a predetermined action of a robot for collecting additional evidences; and a probabilistic information fusion unit for estimating a location and a pose of the object through particle filtering of the various generated locations and poses and expressing the estimated location and pose in arbitrary distribution of particles.
-
Specification