Sensors and Systems
Keith Halcro, MSc
University of Saskatchewan
Disclosure: Disclosure information not submitted.
Lentil seed characteristics are important for marketing, and have typically been evaluated visually and on a bulk-sample basis. BELT (Better Evaluation of Lentils Technology) is a hopper-fed system that singulates seeds onto a moving belt and passes the seeds through an imaging chamber illuminated by high colour rendering index 6500K LEDs. BELT captured 24 bit-depth RGB PNG images which held a top view and side view of each seed, reflected through a 45-degree prism. Images were fed into phenoSEED, a Python3 image analysis script. phenoSEED’s preprocessing applied a colour calibration to the images and segmented the top and side views. Colour calibration was achieved by applying a trained artificial neural network to transform each pixel from RGB to L*a*b* colour space. The preprocessed items include a colour-corrected L*a*b* colour space image and binary mask images that represent top and side shape of the seed. Preprocessed results are saved to disk, then reloaded into phenoSEED which extracts information such as diameter, height or colour distribution across the seed coat.
BELT and phenoSEED have been used to extract information about lentil seeds from breeding trials to obtain objective and repeatable measurements of traits. Traits such as diameter, height, and colour distribution were measured on a per seed basis and aggregated in statistical comparisons of lentil populations to their parent lines. Using BELT was forecasted to eliminate operator biases, increase the rate of acquisition of quantitative traits and branch into assigning qualitative traits that have historically been analyzed by eye. BELT and phenoSEED have been tested with oilseeds, pulses and cereals, all of which were imaged correctly by BELT and segmented in phenoSEED. Possible expansions for BELT include refinements for real-time seed sorting, adding optical coherence tomography to examine subsurface elements or spectral cameras to explore detailed colour spectra.