博文

week4

图片
Week4 22/03/2023 Goals for week4: l    Compare the extracted lens with fitting graph l    Using color map to show deviation l    Calculate the actual surface curvature According to fitting graph, we can get a spherical equat ion: Because we have got scatter plot of the extracted lens, the picture is below that: Then, substitute coordinates of all pairs of points into the spherical equation, which can get deviation of position. And according to the deviation, we can use  color map to show the difference clearly. In actuality, we have come up with many ways to calculate curvature, but some ways are not useful. For example, Gaussian curvature can be calculated by using a 3-D surface where there is equidistant grid variation of points. However, after fitting graph and choosing amounts of deviated points, we only get a set of scatter plot. As a result, we figured out another way to calculate the curvature of scatter plot: fitting a deterministic sphere by selecting   the proximity of a dete

week3

图片
Week3 15/03/2023 Goals: Extraction coordinates Fitting the sphere 1. Extraction coordinates 1) First version: For the original 3D sphere, after slicing: the image pixels are first kept the same size as their pixel matrix and then converted to a heat map: After binarisation: Image after filtering and edge extraction : At this point we find that the extracted edge curves do not accurately represent the position of the lens surface because of the binarization process, and we used a different, more accurate method. 2)  More accurate version Masking : taking the four points on the lens through the luminance curve and substituting them to obtain the equation of the sphere, which is used as a mask to eliminate the image interference of platforms and stray points in the image data. Extraction of coordinates : Based on the masked 3d pixel matrix, get the exact 3d lens coordinate position by taking the point with the largest pixel value in the z-axis pixel curve along the plane formed by the x-a

week2

图片
week2  08/03/2023 Goals: d eveloping the overall idea 1. 3D coordinate position extraction (incomplete) The data are the 3D coordinates and their 3D pixel matrix, sliced along a certain axis we can get a certain 2-dimensional profile. It is then expected to do a binary transformation or greyscale of the image to extract the specific pixel values of the 2D curve and obtain its corresponding 3D coordinates, and so on, for each point along the axis to obtain the coordinates of each point of the 3D surface. Other issues: redundancy culling (interception of pixel matrix), noise reduction. 2. Data analysis Gradual slicing along the x, y and z axes. x axis: Slowly intercepting the profile along the x-axis, the height of the profile arc is found to increase gradually, similar to a hemisphere cross-section, but the arc is found to be asymmetrical Y axis: Similar to x-axis Z axis: Probably a non-standard Hemisphere cross-section from the z axis, while finding the interference of the imaging plat

week1

week1 01.03.2023 Goals: background checks, task checks, organisational arrangements 1. Background Optical coherence tomography (OCT) is a non-invasive and cross-sectional imaging technique that permits, for example, three-dimensional images with micrometer resolution to be obtained from within the retina. Since its invention in 1990, OCT has been adopted as an established quantitative imaging tool for the diagnosis and management of ocular diseases. To date, however, applications of OCT beyond biomedicine have been scarce as most commercially available OCT systems are specifically developed and optimized for medical applications. Recently, we and other groups have demonstrated that OCT can be used for industry applications including non-destructive evaluation of car paints and electronics circuits.  2. Task A typical OCT measurement will yield a 3D data cube with a size of 500x500x1024. In project, you will be provided a set of 3D images (600x550x1600) of spherical and/or aspherical le