SAAIL Lab Publications
Int. Agric. Eng. J. · 2020 Prior Institutional Research · Pre-IITMZ (2020)

Separation of partially occluded rice kernels and variety classification based on computer vision

Tchalla Korohou 1 , Cedric Okinda 1 , Innocent Nyalala 1 , I. Torotwa 1 , E. Mwangi 1
1College of Engineering, Nanjing Agricultural University, Nanjing, China
International Agricultural Engineering Journal, ISSN 0858-2114 (Jan 2020)

Summary: Morphological watershed segmentation and classification for separating touching and occluded rice kernels across grain varieties.

Abstract

Addresses touching and overlapping rice grain clusters using watershed segmentation and morphological shape classification to automate rice variety identification in post-harvest grain inspection.

Keywords: rice kernelsocclusion segmentationvariety classificationcomputer vision

Key Findings & Contributions

Watershed
Occlusion Splitting
Separates touching grains
Morphological
Variety Classifier
Postharvest grain inspection
Multi-Cultivar
Shape + Color Cues
Automated bulk sorting
IAEJ 2020
Published Article
ISSN 0858-2114
  1. Applied distance-transform watershed segmentation to separate touching and partially occluded rice kernels captured in bulk postharvest inspection scenes.
  2. Extracted shape geometry and surface descriptors from separated kernel masks for multi-variety rice classification.
  3. Removed manual single-kernel singulation bottlenecks prior to automated optical grain grading.

Benchmark Results

Grain Separation & Variety Grading Pipeline (IAEJ 2020)

Processing StepTechniqueRole in Postharvest Grading
1. Binary MaskingAdaptive thresholding & noise filteringIsolates rice kernel cluster foreground
2. Kernel SeparationDistance transform + Watershed splitSeparates touching/overlapping grains into singles
3. Variety ClassificationGeometric & textural kernel featuresAssigns rice cultivar class without manual sorting

BibTeX

@article{korohou2020ricekernels,
  title={Separation of partially occluded rice kernels and variety classification based on computer vision},
  author={Korohou, Tchalla and Okinda, Cedric and Nyalala, Innocent and Torotwa, I. and Mwangi, E.},
  journal={International Agricultural Engineering Journal},
  year={2020}
}
APA Citation

Korohou, T., Okinda, C., Nyalala, I., Torotwa, I., & Mwangi, E. (2020). Separation of partially occluded rice kernels and variety classification based on computer vision. International Agricultural Engineering Journal.