This is a design and engineering project, to develop a 3D printed ‘light box’ for taking pictures of flat objects, such as leaves, in the field. The image would be taken by a mobile phone. The light box would position and illuminate the leaf in ways that create a clean, evenly but obliquely lit image with no shadows (yet one that can be taken relatively quickly). The phone would fit onto or into the box. Exactly how it is accomplished is up to you!
The larger purpose is provide clean images for a convolution neural network (CNN) trained to do species identification. In order to distinguish the most closely related species, images should be as free as possible from noise caused by things such as uneven illumination, and contain as much structural information as possible, such as the pattern of veining. At the same time they need to be taken quickly, because CNNs require a large amount of images as input (even when using transfer learning). So in the ideal setup it takes no more than a few seconds to insert a leaf and get a good picture.
Suitable for students with a background that includes optics, such as Physics or Engineering students. ‘Maker’ skills (3D printing, component assembly) a plus!
Structural Ecology Lab @ NJIT 