Structure from motion is a photogrammetric technique that is applied across various industries. It works by interpreting 2D images, shape recognition, camera properties, and accurate environment measuring to reconstruct images into a 3D model. This technique takes into account the position of the camera and the object, the number of images taken, the method of recording, and processing to develop a digital output (Papadopoulos, n.d.). By studying the theory and development of this technique through a series of Dariah Teach articles, we had a conceptual foundation that aided the technical process of capturing the model.
Agisoft Metashape is the photogrammetry software tool that we used, it does not have a steep learning curve but rather a series of specific steps that we need to follow. Organization is key when working with any media and software, and this project was no different. First, we established folders for specific image formats. Then we imported the JPEG images into the software workspace. We ran a quality check on the images and disabled every image with a quality below 0.5.
After the workspace is set, we align all the photos being used and then resize the region so it reduces the focus area by excluding the styrofoam or background elements. Next, we built a point cloud, duplicated said point cloud and began to manually tidy up specific points like the outline of Woody's jeans or styrofoam on his cowboy boots. As soon as we were happy with our tidied-up model we built the model and then the texture.
This led to a process where we continued to close holes and smooth in order to refine the model. Refining the model requires retexturing as well. The finished model then generates masks, which can be exported to create a better version of the model. Masks are incredibly useful in image processing as they clean and refine the area of interest for the software to generate accurate results. By using the masks created from the model and in a higher-quality setting, we repeat the process of building a model once more.