.png){kind=icon}
.png){kind=icon}
Return
Model-Based Multiple Rigid Object Detection and Registration in Unstructured Range Data
.png){kind=icon}
PRE
AI
DOI：10.1007/s11263-010-0401-3.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
We present a two-stage approach to the simultaneous detection and registration of multiple instances of industrial 3D objects in unstructured noisy range data. The first non-local processing stage takes all data into account and computes in parallel multiple localizations of the object along with rough pose estimates. The second stage computes accurate registrations for all detected object instances individually by using local optimization. Both stages are designed using advanced numerical techniques, large-scale sparse convex programming, and second-order geometric optimization on the Euclidean manifold, respectively. They complement each other in that conflicting interpretations are resolved through non-local convex processing, followed by accurate non-convex local optimization based on sufficiently good initializations. As input data a sparse point sample of the object's surface is required exclusively. Our experiments focus on industrial applications where multiple 3D object instances are randomly assembled in a bin, occlude each other, and unstructured noisy range data is acquired by a laser scanning device.
Keywords:
Range data
Point set registration
Multiple object detection
Geometric optimization
Sparse convex programming
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
9.3
Papers:
3.9K
Citations:
2.8W