Paper

Title: Identification of wheat kernels by fusion of RGB, SWIR, VNIR samples over feature and image domain Authors: Kemal Özkan, Şahin Işık, Büşra Topsakal Yavuz DOI: https://doi.org/10.1002/jsfa.9732

Cite: Özkan, K., Işık, Ş. and Topsakal Yavuz, B., 2019. Identification of wheat kernels by fusion of RGB, SWIR, VNIR samples over feature and image domain. Journal of the Science of Food and Agriculture.

Image Capturing System

• List of 40 Wheat Labels: Ahmetaga, Altay2000, Atay85, Aytin, Bayraktar2000, Bezostaja, Cesit1252, Cetiner, Dagdas, Ekiz, Energo, Es26, Esperia, Flamura85, Gelibolu, Gerek79, Harmankaya, Izgi, KateA1, Kirac, Kirgiz, Kunduru, Mesut, Michelangelo, Midas, Mufitbey, Nacibey, Pehlivan, Porsuk, Reis, Renan, Selimiye, Sonmez, Soyer, Sultan95, Suzen, Tosunbey, Yelken, Yildiz, Yunus.

Visible-Near Infrared (VNIR) Wheat Dataset

• The dataset is constructed with a VNIR camera that is able to capture the spectral information between wavelength range of 400 nm to 1100 nm.
• There are 40 wheat classes and 200 samples per each class.
• A captured image is in the form of 640 x 512 pixels.
• Then, each image is divided into four quadrants for data augmentation.
• Two 100-W halogen light sources are used by positioning at the left side of SWIR and right side of RGB to ensure homogeneous light distribution.
• Download VNIR Dataset

Short Wave Infrared (SWIR) Wheat Dataset

• The dataset is constructed with a SWIR camera that is able to capture the spectral information between wavelength range of 900 nm to 1700 nm.
• There are 40 wheat classes and 200 samples per each class.
• A captured image is in the form of 640 x 512 pixels.
• Then, each image is divided into four quadrants for data augmentation.
• Two 100-W halogen light sources are used by positioning at the left side of SWIR and right side of RGB to ensure homogeneous light distribution.
• Download SWIR Dataset

RGB Wheat Dataset

• The RGB images are captured with a specific camera.
• There are 40 wheat classes and 200 samples per each class.
• A captured image is in the form of 1280 x 1024 pixels.
• Then, each image is divided into four quadrants for data augmentation.
• Two 100-W halogen light sources are used by positioning at the left side of SWIR and right side of RGB to ensure homogeneous light distribution.
• Download RGB Dataset

Results

Coming.

License

Free only for research and educational purposes, for commercial use, please contact to corresponding author.

Cite

Özkan, K., Işık, Ş. and Topsakal Yavuz, B., 2019. Identification of wheat kernels by fusion of RGB, SWIR, VNIR samples over feature and image domain. Journal of the Science of Food and Agriculture.

Contact

• Dr. Şahin Işık - Corresponding Author - Şahin Işık - Email - [email protected]