VPBR: An Automatic and Low-Cost Vision-Based Biophysical Properties Recognition Pipeline for Pumpkin

Area of Science:

• Agricultural Science
• Computer Vision
• Biophysics

Background:

• Pumpkin yield is influenced by biophysical properties, crucial for agricultural productivity.
• Manual measurement of these properties is labor-intensive and time-consuming.
• Developing automated methods is essential for efficient crop monitoring.

Purpose of the Study:

• To develop and validate an automated system for measuring pumpkin biophysical properties using high-resolution imagery.
• To compare the accuracy and efficiency of the automated system against manual measurement techniques.

Main Methods:

• Utilized high-resolution pumpkin images for training a mathematical model.
• Implemented color correction using a color-checker panel to standardize image data.
• Trained a segmentation model to differentiate pumpkin fruit and vine components.
• Validated measurements of fruit length, width, stem length, stem width, and peel color against manual data.

Main Results:

• The automated system achieved a low average mean absolute percentage error (MAPE) of 2.5% compared to manual measurements.
• Demonstrated successful segmentation of pumpkin fruit and vine.
• Validated the measurement of key biophysical properties.

Conclusions:

• The image-based automated system provides a faster and more efficient method for monitoring pumpkin biophysical properties.
• This technology has the potential to significantly improve agricultural practices in pumpkin cultivation.
• Further research can expand this approach to other crops and biophysical parameters.