WISRD Research & Engineering Journal

Figure 38 shows the HSV (hue, saturation, value) color spaces across the top row of images and the Lab (“L” channel is lightness, the “a” channel is green-magenta, the “b” channel is blue- yellow) across the bottom row.

Figure 38 . Conversion of RGB photo to HSV and Lab color spaces. In certain channels, the target object will either be lighter or darker than the background. A threshold is applied whereby, for example, anything above the threshold is set to white and anything below the threshold is set to black. The image was first passed through the “b” channel, a tentative threshold was set, and any pixels above the threshold were set to 255 for white and anything below was set to 0 for black. The image was also passed through the “a” and “s” channels, and then the results of the three channels were combined. There was still some noise (“salt and pepper’ effect) that needed to be removed and this was done by a filling step whereby anything smaller than a defined number of pixels gets covered up. The process of segmentation is shown in Figures 39-42.

Figure 39 . Generating color spaces is the first step in segmenting photos using PlantCV.

57