Size measurement and filled/unfilled detection of rice grains using backlight image processing

Area of Science:

• Agricultural Science
• Computer Vision
• Biotechnology

Background:

• Accurate measurement of rice physical traits is vital for effective crop breeding programs.
• Traditional methods for rice grain analysis can be time-consuming, destructive, and costly.
• Developing objective and efficient phenotyping tools is a key objective in modern agriculture.

Purpose of the Study:

• To present a novel, non-destructive image processing approach for quantifying rice physical traits.
• To assess the accuracy and efficiency of the proposed method for grain analysis.
• To investigate the influence of backlight intensity on measurement precision.

Main Methods:

• Utilized backlight photography to capture grayscale images of rice grain samples.
• Applied a clustering algorithm to differentiate filled and unfilled grains based on grayscale values.
• Quantified grain dimensions (length, width) and particle counts using image analysis.

Main Results:

• Achieved high accuracy with a mean absolute percentage error of 0.24% for total grain count and 1.36% for filled grain differentiation.
• Demonstrated low error margins for grain size measurements: 1.11% for length and 4.03% for width.
• The method proved to be highly accurate, efficient, non-destructive, and cost-effective.

Conclusions:

• The proposed image processing technique offers a precise and efficient solution for rice grain phenotyping.
• This non-destructive, cost-effective method is a valuable tool for accelerating rice breeding programs.
• The findings support the adoption of advanced imaging technologies in agricultural research and development.