Detecting and Localising Multiple 3D Objects: A Fast and Scalable Approach

Research output: Chapter in Book/Report/Conference proceedingConference Contribution (Conference Proceeding)

215 Downloads (Pure)


Object detection in complex and cluttered en-
vironments is central to a number of robotic and cognitive
computing tasks. This work presents a generic, scalable and fast
framework for concurrently searching multiple rigid texture-
minimal objects using 2D image edgelet constellations. The
method is also extended to exploit depth information for better
clutter removal. Scalability is achieved by using indexing of a
database of edgelet configurations shared among objects, and
speed efficiency is obtained through the use of fixed paths which
make the search tractable. The technique can handle levels of
clutter of up to 70% of the edge pixels when operating within
a few tens of milliseconds, and can give good detection rates.
By aligning our detection within 3D point clouds, segmentation
and object pose estimation within a cluttered scene is possible.
Results of experiments on the challenging case of multiple
texture-minimal objects demonstrate good performance and
scalability in the presence of partial occlusions and viewpoint
Original languageEnglish
Title of host publicationIEEE IROS Workshop on Active Semantic Perception and Object Search in the Real World (ASP-AVS-11)
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
Publication statusPublished - Oct 2011


Dive into the research topics of 'Detecting and Localising Multiple 3D Objects: A Fast and Scalable Approach'. Together they form a unique fingerprint.

Cite this