This study proceeds with a systematic review and subsequent comparison between the geometry-based and regression model-based techniques on accurately estimating the physical properties of avocados particularly mass using manually collected cross-sectional data set. The objectives of this investigation are to handle grading, weighing and packaging effectively, and cost-effectively for the agricultural and food engineering industries. Among the findings, the Frustum method consistently outperforms other approaches across all slice configurations, achieving the lowest errors with Root Mean Square Percentage Error (RMSPE) of 4.24%, Mean Absolute Percentage Error (MAPE) of 3.43%, Coefficient of Determination () value of 97.78% and Explained Variance Score (EVS) value of 97.63% at 20 slices. This highlights its robustness and reliability for precise mass estimation, especially without requiring large datasets or complex computations. To ensure the reliability of the regression models, hyper-parameter optimization and K-fold cross-validation techniques were employed, enabling the identification of optimal model configurations and minimizing over-fitting risks. Regression-based methods, such as Ridge Regression, also exhibit strong performance, with an average RMSPE of 4.30%, MAPE of 3.52%, of 97.71%, and EVS of 97.58% across 5 folds at 15 slices, making it a competitive and stable alternative. Other regression models, such as LASSO Regression, and Elastic Net Regression, delivered strong and consistent outcomes across the evaluation metrics, followed by Linear Regression. In contrast, MLP Regressor and Gradient Boosting Regressor exhibited notable variability between folds, highlighting issues with stability and generalization. Dimensions of avocados were also assessed in this study with the error rating below 1.53% coupled with model-fit parameters of more than 99% showing that the models used had high accuracy in determining both the width and length of the avocado. The above findings provide a comparative perspective of the choice of forward models depending on the task characteristics including, availability of dataset, stability, and level of precision. Moreover, such outcomes reveal the applicability of these methodologies in implementing state-of-the-art automation technologies accordingly with a focus on robotic harvesting and grading solutions in automation of precision agriculture and modern intelligent food processing systems.

Avocado (Persea americana Mill.) fruit is a desirable plant in economical and health sectors in tropical and subtropical areas. The avocado is a rich source of lipids, vitamins, minerals and bio-active compounds and as such an essential part of human diets and food industries the worldwide. Global avocado production has steadily increased over the past decades, with major producers including Mexico, Colombia, Peru, Dominican Republic, Indonesia, and others in Asia. Traditionally, sorting and grading of avocado after harvest was still done using manual methods which included visual way of determining size, shape, and color or directly reading sizes using calipers and weighing scales. Even though these are cheap and easy, they take a long time, are subjective, labor intensive and can be easily spoiled by human error, especially in the case of large amounts of fruits. Besides, manual sorting is inconsistent in terms of throughput that supply chains demand. Such constraints have prompted the use of modern technologies such as computer vision, image analysis, and machine learning that will enable non-destructive, objective and high-throughput measurement assessment of fruit characteristics. These strategies not only make postharvest operations more accurate, but also enhances automation in precision agriculture.

Fruit mass and physical characteristics estimation is one of the core components of agricultural technology incorporating the desired goals of increasing yields, efficiency, and quality, and decreasing costs in this route. Growing food necessities across the globe on account of the ever-growing population makes efficiency immaculate, making use of precision agriculture and automation inevitable. They allow producers to manage resources effectively, fulfill customers' requirements, and conform to industry standards. Among all parameters involved, density and components like size, form, and surface area are important for the identification of quality, easy evaluation of quality, and solutions in further procedures like sorting, grading, and packaging.

Prior work using related technology areas such as computer vision and image processing has remained challenging but efficient in estimating such important physical features. For example, Lorestani and Tabatabaeefar (2006) showed how length, width, and height could be used in models to determine the mass and volume of kiwis. According to their results, is up to 0.97, explaining the coefficient of determination, meaning that their approach for mass estimation is rather accurate. Thus Emadi et al. (2009) considered the mechanical characteristics of melons, including the compression and shear stresses, for the study of size and shape characteristics. Not only did their study offer correct classification (all varieties were statistically similarly classified with p > 0.05), but their work also aided applications in optimizing automated grading systems through the incorporation of such mechanical properties into the grading assessment by machines. Further work expanded these methods to other fruit types incorporating both visual and tangible qualities. Liming and Yanchao (2010) proposed a non-destructive method to model the surface area and symmetry of strawberries for proper packing and aesthetic purposes. Their research revealed high accuracy Levels that in turn helped improve the method of sorting and packaging. The data they captured for classification and color grading analysis was more than 90% with an object size detection error of not more than 5%. The methods developed by Clayton et al. (1995) were aimed at estimating the surface area of apples - rather important for ripening behavior and shelf-life prediction. As it was noted, their model has provided outfits with an increased precision rate effective for managing inventory processes. These results described the significance of consumer-sensitive characteristics such as symmetry in the post-harvest processes while also capturing concerns such as fruit surface area in the internal decision-making processes.

More recently, developments in technology have improved the efficiency of these estimation techniques. Sripaurya et al. (2021) have used the portable non-contact 6 digital channel near-infrared to measure soluble solids and maturity of banana. Their system not only precisely estimates size and shape but also well classified seven ripeness Level, standard error of the prediction is 0.17% and standard error of the stages of color is 0.38%. Likewise, Chen et al. (2022), propose a new method using the convolutional neural network to detect citrus fruit maturity for selective robotic picking due to the fact that citrus fruits on trees are at different ripening stages. Fruit detection is done using YOLOv5, while maturity Levels are classified in a 4-channel ResNet34 that feeds both saliency maps and RGB images. Their proposed approach was able to capture an accuracy of 95.07% which is way better than the VGG16 model as well as KNN, by having 3.14%, 18.24% better accuracies respectively. Oyefeso et al. (2018) presented mathematical models to predict the mass and volume of sweet and Irish potatoes using geometry in 2018. The study also found that both of the most accurate models developed for the crops were derived from projected area and volume where coefficients of determination of 98.10% were obtained for sweet potatoes and 96.80% for Irish potatoes. Such findings are useful for planning effective means of handling and processing these tuber crops in a more orderly manner. A novel Intelligent Flexible Manipulator System with Flexible Tactile Sensing (IFMSFTS) was proposed in 2023 by Qin, et al. In particular, kiwifruit has been losing up to 25% during its post-harvesting process throughout the world. This system employs a flexible manipulator with a force sensor to evaluate the firmness and associates firmness and ripeness to categorize the kiwifruit maturity levels. Hence, applied Principal Component Analysis (PCA) to reduce the data dimension and used two classifiers, K- nearest neighbor (KNN) and support vector machine (SVM) for training and testing with classification rates of 97.5% and 96.24%, respectively while KNN showed slightly better performance than SVM. This technology is ideal for classifying ripeness and minimizing fruit losses hence encouraging sustainable kiwifruit production.

However, there are still difficulties in applying these techniques in large scale production lines, which are preferred in high-throughput systems. The sophisticated planning methods analyzed in this survey tend to deliver high accuracy at the price of slow computations, which sharply constrains the application of these systems in the dynamic context of agriculture. While such approaches have been tested to yield good outcomes mainly in a laboratory environment, practical problems need solutions that are fast, accurate and affordable. Further, these limitations are crucial to avoid and promote better performance of automated systems that could estimate fruit mass and physical characteristic under various circumstances.

Mass estimation, in particular, is vital for tasks such as sorting, grading, and packaging, which are integral to maintaining consistent quality and profitability. In contrast to ripeness detection and quality grading, little is known about the methods to estimate the mass of fruits non-destructively, quickly, and on a large scale.

Hieu M. Tran et al. (2023) presented a series of studies addressing the issue of segmenting the shape of irregular fruits, particularly focusing on starfruit (Averrhoa carambola) from viewpoint of dimensions, volume, and mass. These physical characteristics are important for the design of grading, sizing, and packaging equipment and aspects within the agricultural industry. Traditional approaches are known to encounter problems while working with objects of complex shapes; several measurements or complex three-dimensional modeling may be needed. Thus, the authors of the Tran et al. work proposed innovations based on image processing and simple mathematical algorithms, which proved high accuracy, yet remained plausible for implementation. These two research pieces focus on the development of solutions that bypass the traditional shortcomings in volume measurement For scale accuracy, almost perfect results are obtained for volume and mass estimation, particularly for starfruit and similar fruits in industrial application.

The first study by Tran et al. used stylized photographs to directly measure the physical attributes of starfruit using visual data with a dual-camera approach to get top view and body view imagery. Image processing techniques, especially the principal axis of the conical frustum and other mathematical approaches were used to virtually slice the star fruit along its major diametrical line to arrive at the volume. The mass was then estimated by integrating the volume with the fruit density owing to the density's proportionality to the mass. The applicability of the described method was tested using 255 training samples, providing the overall mean accuracy of the volumes of 99.16% and the mass of 98.59%. This approach clearly illustrated why simplicity and reconfigurable are advantageous in manufacturing environments and shows how it can form a solid solution for gauge development to measure round fruits. In the second investigation, Tran et al. brought down the complication of the system by using a single camera, although they paid a high accuracy price. The process was divided into two phases: to start with, a primary perspective photograph was taken, and picture manipulative procedures were applied to dissect the fruit in the sagittal line. The volumes of each disc were estimated using the disc and conical frustum methods where the summation of these yielded the total volume. In the second phase, the linear regression coefficient of determination = 0.9205 was used to predict the mass in terms of volume. The simplified flow chart successfully passed 300 samples of testing, which gave 99% accuracy for volume and mass, making this a simple and usable method for measurements in stars fruit for its industrial use.

In addition, in 2022 Huynh et al. designed a single top view camera vision based simple system to estimate sweet potato attributes with geometrical Transformations in which the product can be virtually sliced following its longitudinal plane and the total volume can be obtained from the sum of all slice volume. The degree of relationship between volume and weight is high and therefore provides an excellent measure of mass, an accuracy of 96% for volume and 95% for weight. The system is faster and does not require computationally heavy 3D reconstructions and makes it practical and cost effective for real life applications.

Although geometric and statistical methods to solve the weight prediction of agricultural produce have recently been advanced, work specifically dealing with the avocado fruit is scarce. Previous investigations also limited the predictive results on simple geometric variables, which limits the scope of the results and applicability at large-scale production. To bridge this gap, the present study suggests two complementary solutions to mass determination of avocados, which are geometry-based frustum approach and machine learning regression working with morphological features and image-based attributes. The goal is to enhance the precision and strength of weight prediction of avocadoes whilst doing so in a manner that is feasible to practical farming. This research problem is addressed with the following research questions: (1) How well can geometric and image-based characteristics be used to improve mass prediction accuracy over the traditional prediction methods? (2) Which regression algorithms give the best result of avocado weight estimation? (3) What are the proposed methods contribution to the automation opportunities, e.g. vision-based weighing systems, or robotic harvesting systems? To address these questions, Section 2 describes the materials and methods, where the measurement techniques, model frameworks, validation strategies, and relevant formulas are detailed. Section 3 reports the results and discussion, including performance analysis of the proposed approaches and comparisons with existing studies. Finally, Section 4 concludes with the main findings, implications for agricultural automation, and future research directions.