3D Object Detection and Tracking Based On Point Cloud Li- brary Special Application In Pallet Picking For Autonomous Mobile Machines
Estiri, Fatemeh Alsadat (2014)
2014
Master's Degree Programme in Machine Automation
Teknisten tieteiden tiedekunta - Faculty of Engineering Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2014-05-07
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201405131168
https://urn.fi/URN:NBN:fi:tty-201405131168
Tiivistelmä
This work covers the problem of object recognition and pose estimation in a point cloud data structure, using PCL (Point Cloud Library). The result of the computation will be used for mobile machine pallet picking purposes, but it can also be applied to any context that requires finding and aligning a specific pattern.
The goal is to align an object model to the visible instances of it in an input cloud. The algorithm that will be presented is based on local geometry descriptors that are computed on a set of uniform key points of the point clouds. Correspondences (best matches) between such features will be filtered and from this data comes a rough alignment that will be refined by ICP algorithm. Robust dedicated validation functions will guide the entire process with a greedy approach. Time and effectiveness will be discussed, since the target industrial application imposes strict constraints of performance and robustness.
The result of the proposed solution is really appreciable, since the algorithm is able to recognize present objects, with a minimal percentage of false negatives and an almost zero false positives rate. Experiments have been conducted on datasets acquired from a state-of-the-art simulator and some sample scene from the real environment.
The goal is to align an object model to the visible instances of it in an input cloud. The algorithm that will be presented is based on local geometry descriptors that are computed on a set of uniform key points of the point clouds. Correspondences (best matches) between such features will be filtered and from this data comes a rough alignment that will be refined by ICP algorithm. Robust dedicated validation functions will guide the entire process with a greedy approach. Time and effectiveness will be discussed, since the target industrial application imposes strict constraints of performance and robustness.
The result of the proposed solution is really appreciable, since the algorithm is able to recognize present objects, with a minimal percentage of false negatives and an almost zero false positives rate. Experiments have been conducted on datasets acquired from a state-of-the-art simulator and some sample scene from the real environment.