Intelligent Fruit Yield Estimation for Orchards Using Deep Learning Based Semantic Segmentation Techniques-A Review.

Area of Science:

• Agricultural Technology
• Computer Science
• Artificial Intelligence

Background:

• Smart farming utilizes advanced technologies for sustainable agriculture and economic growth.
• Manual fruit yield estimation is labor-intensive, time-consuming, and often imprecise.
• Deep Learning (DL) offers state-of-the-art results in smart farming applications, particularly for yield estimation.

Purpose of the Study:

• To review literature on Deep Learning-based semantic segmentation for fruit yield estimation.
• To identify challenges in intelligent fruit yield estimation.
• To discuss future directions for enhanced yield prediction systems.

Main Methods:

• Review of various Deep Learning semantic segmentation architectures used in fruit yield estimation.
• Analysis of challenges including data collection, annotation, augmentation, fruit detection, and counting.
• Examination of DL's pixel-based prediction for fruit localization and detection.

Main Results:

• Deep Learning-based semantic segmentation significantly outperforms traditional methods in fruit yield estimation.
• The incorporation of human-like cognition into DL architectures improves prediction accuracy.
• DL techniques provide more precise fruit detection and localization.

Conclusions:

• Deep Learning semantic segmentation is a powerful tool for intelligent fruit yield estimation in smart farming.
• Addressing challenges like occlusion, overlapping, and illumination variation is crucial for future models.
• Customizing DL for smartphone applications can further benefit farmers.