Color image segmentation approach to monitor flowering in lesquerella

Lesquerella (Lesquerella fendleri (Gray) Wats.) seed oil has been proposed as a petroleum alternative in the production of many industrial products, but several crop management and breeding challenges must be addressed before the crop will be grown commercially. Lesquerella canopies characteristically exhibit quite prominent and vibrant yellow flowers at anthesis, and remote detection of lesquerella flowering patterns can provide useful crop development information to aid management and breeding decisions. In the present study, we used a consumer-grade digital camera to collect images 2 m above lesquerella canopies throughout two growing seasons. Biomass samples within 0.125 m(2) areas were also regularly collected and processed to obtain flower numbers. Image processing algorithms were developed to extract information on lesquerella flower features from the images. Key features of the image processing approach included an image transformation to the hue, saturation, and intensity (HSI) color space and a Monte Carlo approach to address uncertainty in HSI parameters used for image segmentation. Flower numbers were estimated from image-based flower cover percentage with root mean squared errors that ranged from 159 to 194 flowers, which was better than the reported results for other studies with a similar objective. Attempts to resolve individual flowers were less successful due to the complexity of the flowering patterns within the image scenes. Digital imaging offers an inexpensive and quite practical means for remote monitoring of flowering patterns in lesquerella canopies. Published by Elsevier B.V.