stevengoes / ssic Goto Github PK

Several apps for 3D scanning exists but most of them need extra hardware or calibration tools (link). Compared to other photogrammetry methods, Shape-from-Silhouette (SfS) is computationally simple (paper) and the scanning procedure can be performed marker-less (paper). However, the digital model from SfS is often an inaccurate representation of the real shape. These SfS models are also known as visual hull (VH). But the availability of large-scale anthropometric surveys using 3D body scanning technologies has enabled researchers to predict 3D body shapes from partial inputs, such as a small set of silhouette pictures (paper),(paper). A statistical shape model (SSM) is designed with the survey models and can be used to morph and fit within the object outline in the silhouette image. This silhouette-based shape estimation will be called the statistical shape image comparison (SSIC) method. It compares projections of the SSM with silhouette images to validate and morph the SSM into the desired 3D shape. This technique has brought potential benefits to the e-commerce of wearable products but might just as well be used in the medical field for wearable devices such as prostheses.

This research is part of a TU Delft Global Initiative project named ”Access to prosthetics thanks to 3D printing and a smartphone app” (link). The goal of the project is to make prosthetic devices easy accessible by automating the design process. A prosthetic hand model with no need for assembly, a parametric socket design, cheap materials for 3DP and a smartphone as a residual limb measuring tool will enable a fast and low-cost production. The process will be an end-to-end solution where the patient will scan his/her residual limb as input and a fitting prosthesis will be delivered as output. The project is meant particularly for Colombia.

This Github repository provides code for setting up a testenvironment to test the possibilities of the SSIC method, without using reallife subjects/objects. The code is in my case specifically used for the 3D reconstruction of a residual limb for prosthetic designing. But the SSIC testenvironment can also be used for other objects as long as you can provide your own database of shapes. In case you have your own databse, the file point_set_registration.m can make the database coherent which is needed for the SSIC method to function. When a registered databse is readily available, the following process can be executed:

The main files need for this process are:

1. ImageAcqBlender.py
2. SSIC.m
3. silhouetteERROR.m

Provide code examples and explanations of how to get the project.

Describe and show how to run the tests with code examples.

• There is still something wrong with Ncam=1. The P matrix is calculated incorrectly in Blender. This problem is under investigation. [update] The RTmatrix is calculate incorrectly for some reason.

Let people know how they can dive into the project, include important links to things like issue trackers, irc, twitter accounts if applicable.

MIT license