Computer Vision for Chrolophyll Level Estimation
Jul 19, 2020
Part of the UTEC-SEAS 7th Collaborative Field Program in Peru; a collaboration between Harvard School of Engineering and Applied Sciences (SEAS) and the University of Engineering and Technology (UTEC).This project started as the pandemic emerged, so we had to replace the original objective with one that could be achieved from home. Hence, we decided that it made sense to try to obtain accurate measurements of chlorophyll levels from phone images alone, since we lacked any other tools. We used object detection algorithms to extract plants out of an image, as well as reference color patterns to adjust for camara filters. After adjusting colors, we used k-means clustering to extract the most representative color out of each plant, which we later used to estimate the chlorophyll level using a formula.
Image of a chrolophyll sample. Taps serve as references for color adjustment.
Object detection with OpenCV to find taps
Choosing the color of any object found in the sample.
Assigning objects into known colors clusters.
Color normalization between two samples for comparable target chlorophyll color.