Active exploration and keypoint clustering for object recognition

Abstract

Abstract-Object recognition is a challenging problem for artificial systems. This is especially true for objects that are placed in cluttered and uncontrolled environments. To challenge this problem, we discuss an active approach to object recognition. Instead of passively observing objects, we use a robot to actively explore the objects. This enables the system to learn objects from different viewpoints and to actively select viewpoints for optimal recognition. Active vision furthermore simplifies the segmentation of the object from its background. As the basis for object recognition we use the Scale Invariant Feature Transform (SIFT). SIFT has been a successful method for image representation. However, a known drawback of SIFT is that the computational complexity of the algorithm increases with the number of keypoints. We discuss a growing-whenrequired (GWR) network for efficient clustering of the keypoints. The results show successful learning of 3D objects in real-world environments. The active approach is successful in separating the object from its cluttered background, and the active selection of viewpoint further increases the performance. Moreover, the GWR-network strongly reduces the number of keypoints.

A human-assisted approach for a mobile robot to learn 3D object models using active vision

“…Fully automatic object segmentation is also possible, using active exploration [13], [14]. There remain some constraints on the objects and the environment: [13] is unsuitable for large objects and for both [13], [14] the object needs to be placed in front of the robot.…”

“…There remain some constraints on the objects and the environment: [13] is unsuitable for large objects and for both [13], [14] the object needs to be placed in front of the robot. Furthermore, an automatically created segmentation may not necessarily correspond to how a human perceives object segmentation.…”

“…Furthermore, an automatically created segmentation may not necessarily correspond to how a human perceives object segmentation. Like Kootstra et al [14] we feel that it is very important for robots to have the ability to actively explore any environment and dynamically update their models accordingly, and we have therefor adopted their methods.…”

“…To overcome such issues, object models based on invariant features such as SIFT [18] and SURF [19] can be used. In [14] SIFT is used in combination with active vision, to create an object model which holds different views of an object. In our approach, we use SURF feature points as SURF is faster and more robust than SIFT [19], we also store the 3D position of these points as computed by a stereovision system.…”

“…In our approach, we use SURF feature points as SURF is faster and more robust than SIFT [19], we also store the 3D position of these points as computed by a stereovision system. By using a human teacher we can loosen the restraint placed on the environment in other works on active vision [14], [13], while keeping the advantages, such as recognizing objects from multiple poses.…”