Large-Scale Counting and Localization of Pineapple Inflorescence Through Deep Density-Estimation

Area of Science:

• Agricultural Science
• Computer Vision
• Remote Sensing

Background:

• Natural flowering in pineapples impacts fruit quality and harvest timing.
• Manual crop inspection is limited by high planting densities, hindering accurate field status assessment.
• Synchronized harvesting is crucial for pineapple production efficiency, cost reduction, and waste minimization.

Purpose of the Study:

• To develop an automated system for counting flowering pineapple plants using computer vision.
• To enable growers to optimize management practices through accurate crop intelligence.
• To improve the efficiency and reduce losses in pineapple harvesting.

Main Methods:

• A deep learning-based density estimation approach was employed for inflorescence counting.
• The method was tested for accuracy, achieving a Mean Absolute Error (MAE) of 11.5 and Mean Absolute Percentage Deviation (MAPD) of 6.37%.
• An active learning framework was integrated for continuous model improvement and learning.

Main Results:

• The deep learning model accurately counted flowering pineapple plants in a field.
• The method demonstrated scalability, efficiently detecting over 1.6 million flowering plants.
• Computational complexity was independent of plant number, ensuring efficient processing.

Conclusions:

• Automated inflorescence counting using deep learning enhances crop management for pineapple growers.
• This technology facilitates optimized harvesting, leading to improved fruit quality and reduced economic losses.
• The developed system offers a scalable and efficient solution for large-scale pineapple field monitoring.