Fruit Weight Prediction and Cost Estimation Using YOLOv9-based Deep Learning

Accurate fruit weight and ripeness measurements play a crucial role in optimizing yield and maximizing profits in agriculture. This paper introduces a methodology that uses image recognition to detect fruit weights and deep learning-based YOLOv9 to classify the level of ripeness. The real-world spatial scales for fruit dimensions are calculated by extracting Exchangable Image File Format (EXIF) metadata from captured images. The proposed approach uses an estimation of fruit weight from specific height factors and fruit density based on the Ground Sampling Distance (GSD) and real-world area to predict weight. A module that estimates costs analyzes prices across regions to recommend the most profitable selling points based on transportation costs. Experimental results show high precision of over 95% accuracy in estimating both weight and ripeness classification, such that farmers can use this insight to their advantage for yield forecasting and market optimization. This integrated approach applies computer vision and market analytics using advanced modes to boost agricultural productivity and economic returns.