Solution to a problem pitched by Normal Warthmann at the 2016 Canberra HealthHack. Solution by group: seedid: Norman Warthmann Thomas Hamer Allen Huang Joseph Meltzer Manal Mohania William Shen
Images will be taken of seed samples, along with the seed bag, a ruler, and a 24 colour sample card. From these images, the seeds within the sample must be separated and analysed. The analysis should include measurements of the length and width of each seed, and summary statistics provided.
In order to standardise and aid image segmentation, all images will be normalised to a certain brightness. The pure white square on the colour card will be detected, and the entire image's brightness will be adjusted until the white square matches a pre-determined brightness level.
Different photographs all have varying scales, depending on camera position and resolution. The 24 colour card is isolated and its length measured, and then a scale factor is calculated and stored, to be used in calculations at the end.
The seeds within the image are all detected, and each individual is isolated into a separate image. ........
The individual images are processed, and the length and width of the seeds inside are determined. These metrics are used to work out if the image contained a single seed or not, and information from images that contained multiple seeds is deleted.
With the data collected from single seed images, tables and histograms are produced, and displayed to the user.