| 研究実績の概要 |
This year focused on the development of a paper-folding machine. Previous prototype (LEGO-based) was redesigned (under construction) using optical sensors, step motors and a PSoc4 kit to eliminate errors (backslash) detected with LEGO motors. The algorithms ware re-programmed with the PSoC4 kit to create a pattern from a 3D object that can be represented by a surface in revolution (SR). The method was presented at national [11-DD, 2-JSST] and international conferences [5-ASME, 13-ICMMA], mass media [15-NHK, 16-TokyoTV] and published in proceedings [1-ASME, 2-JSST]. However, as SR cannot represent the bones accurately the method was extended to create patterns of 3D objects that fulfil the condition that their axial projections in 2d are star-based polygons. Additionally we improved the algorithms for 3D reconstruction of X-ray images and developed machine-learning algorithms based on deep neural networks (DNN) to recognize different scenes (e.g. folding, gluing, sharpening) of an origami process. XVIS toolbox was used to generate a large amount of x-ray images from a bone database and evaluate reconstructed 3D models objectively. In addition to origami videos, we worked with complex time series such as car scenes, facial expressions and data from patients with diabetes. DNNs were applied for autonomous driving, kansei evaluation and predicting human behavior in combination with fuzzy and rough sets. The papers were also presented at national [3-jsst, 6-CMD, 7-CMD, 8-CMD, 9-JSST, 14-Meiji-Marianna] and international conferences [4-ISFUROS, 12-ICMMA].
|
| 今後の研究の推進方策 |
As originally planned, in the following year we will work on two fundamental lines: 1) Generalization of the method to more complex forms and 2) Adaptation of the system to clinical conditions
1) Generalization of the method to more complex forms: Although the developed method can work with complex forms such as bones, it is necessary to develop a segmentation algorithm of the 3D models so that all the parts to be represented in the pattern comply with the condition that their projections in 2d are "star-based polygons" and also that the number of bends and areas of bonding are minimal.
2) Adaptation of the system to the clinical conditions: For the introduction of the system in the clinic it is necessary to develop a prototype of software that allows to create 3D models from X-ray images in an optimal way, by means of the selection of an ROI in each image avoiding drawing the outline of the bones that depends on the pressure of the wrist and generates errors. After obtaining the 3D models and the patterns to reproduce the models, the simulation was used to take the Norigami pattern to the 3D form but the effects of the paste in the simulation are still to be included.
|
