I'm not an expert with OpenCV but I think it possible to train it to recognized object. We already had some recognition algorythm for face tracking and face recognition and have object tracking.
However, to reach your goal you need more than object recognition from openCV. OpenCV gives a 2D perspective, wich is good to have the head look in a direction. But to have the arm take an object, you also need to have the depth data and have a 3d perspective.
I'm currently trying to use the kinect to locate objects in the surrounding of the robots and avoid or reach those object while the arms are moving.
Other alternative could be to use multiple camera to deduce the location of the object, or some other distance sensor like a lidar.
I'm not an expert with OpenCV
I'm not an expert with OpenCV but I think it possible to train it to recognized object. We already had some recognition algorythm for face tracking and face recognition and have object tracking.
However, to reach your goal you need more than object recognition from openCV. OpenCV gives a 2D perspective, wich is good to have the head look in a direction. But to have the arm take an object, you also need to have the depth data and have a 3d perspective.
I'm currently trying to use the kinect to locate objects in the surrounding of the robots and avoid or reach those object while the arms are moving.
Other alternative could be to use multiple camera to deduce the location of the object, or some other distance sensor like a lidar.
Used kinect is a very good
Used kinect is a very good idea. Actually, OpenCV can use the kinect with MRL ?
Thank you