Method for 3D fibre reconstruction on a microrobotic platform

    loading  Checking for direct PDF access through Ovid

Abstract

Summary

Automated handling of a natural fibrous object requires a method for acquiring the three-dimensional geometry of the object, because its dimensions cannot be known beforehand. This paper presents a method for calculating the three-dimensional reconstruction of a paper fibre on a microrobotic platform that contains two microscope cameras. The method is based on detecting curvature changes in the fibre centreline, and using them as the corresponding points between the different views of the images.

Summary

We test the developed method with four fibre samples and compare the results with the references measured with an X-ray microtomography device. We rotate the samples through 16 different orientations on the platform and calculate the three-dimensional reconstruction to test the repeatability of the algorithm and its sensitivity to the orientation of the sample. We also test the noise sensitivity of the algorithm, and record the mismatch rate of the correspondences provided.

Summary

We use the iterative closest point algorithm to align the measured three-dimensional reconstructions with the references. The average point-to-point distances between the reconstructed fibre centrelines and the references are 20–30 μm, and the mismatch rate is low. Given the manipulation tolerance, this shows that the method is well suited to automated fibre grasping. This has also been demonstrated with actual grasping experiments.

Lay Description

The motivation for this paper came from mechanical testing of individual paper fibres, but similar problems arise in testing of other types of fibres as well. The fibres should be fixed from their ends to the measurement system, where stress is applied to the fibres. Manual handling demands dexterity and is time-consuming due to the microscopic size of the fibres. Utilizing test benches composed of microrobotic actuators such as microgrippers in grasping and manipulating the fibres partially overcomes the problem, but it does not remove the need for an experienced operator. The greatest challenge comes from the fact that natural fibres are curly three-dimensional objects. Automated grasping requires a computer-vision-based method to estimate the three-dimensional profile of the fibre in order to calculate sufficient grasping points for the grippers. This paper presents such a method.

Lay Description

We show that the method is able to provide three-dimensional profiles with a high enough accuracy for grasping by comparing the profiles gained with the references measured with an X-ray microtomography device. We also prove that the error rate of the method is relatively low. Moreover, we have successfully used the method in automated grasping experiments.

Related Topics

    loading  Loading Related Articles