부산대학교 도서관

Weight prediction of vegetables by analyzing without using a weight machine instead of using 3D live capturing AI powered trained model by combining (Mask R-CNN) algorithm with Marching Cube Algorithm, voxelization and computer vision technique to find weight. Leveraging advanced computer vision techniques, including the Mask R-CNN algorithm, Marching Cube Algorithm, voxelization, and deep learning models, the study introduces a nonintrusive and efficient method for weight prediction based on 3D representations. The methodology begins with dataset creation, comprising images of vegetables alongside known weights. Mask R-CNN is used for exact object identification, segmentation and to delivering essential ROI information. To build detailed 3D representations of veggies, we use voxelization, which turns segmented 2D ROIs into 3D voxel grids. The Marching Cube Algorithm is then used to convert voxel data into polygonal meshes, allowing for the estimation of surface area, which is an important aspect in measuring vegetable weight. We train a specific weight estimation model using a large dataset that incorporates calculated surface area as a feature. When presented with new vegetable photos, the system performs object identification, voxelization, mesh construction, and surface area estimate. The estimated surface area is utilized to generate an accurate approximation of the vegetable's weight. Continuous validation and modification guarantee that the system is accurate and resilient, taking into consideration real-world elements like vegetable density, wetness, and shape fluctuations. This study proposes a novel, non-intrusive approach for reliably forecasting vegetable weights, with possible uses in agricultural, retail, and food processing.Novelty: This research describes a novel method for determining vegetable weight that does not rely on existing physical weight machines. Vegetables are typically weighed using such machines, which can be sluggish and inefficient, especially for firms dealing with big volumes. Our method is a faster and more exact alternative to using a physical weight machine to measure vegetable weight.