Open Access
Table 2
Literature review on various AI methodologies in food technology.
| Reference | AI methodology | Observations | Remarks |
|---|---|---|---|
| Loisel et al. [15] | Neural networks | This paper discusses the definitions, challenges, and machine learning methods for detecting breaks, highlighting issues and data sources for training models. | Wireless temperature sensors and data transmission giving real-time temperature analysis will enable automatic warning and determination of time-temperature threshold along with break classification. |
| Hoang et al. [16] | Convolutional LSTM model | Implementing demand response requires precision to safeguard food quality and safety. Four Artificial Neural Network (ANN) models, leveraging deep-learning techniques, were devised to anticipate temperature and power demand fluctuations in cold storage resulting from Demand Response applications. | Four deep learning models were developed, including traditional LSTM, stacked LSTM, bidirectional LSTM, and convolutional LSTM. Results show that ANNs can predict product and air temperature evolutions during demand response periods. |
| Khanuja et al. [17] | Auto-Regressive Integrated Moving Average (ARIMA) | This paper suggests a framework that leverages IoT, Cloud Computing, Machine Learning, and Big Data Analytics to transform the cold transport industry, guaranteeing the preservation of freshness and quality. | The IoTs revolutionizes fleet management by utilizing Big Data and Machine Learning to interpret and compute data from sensors. Future work aims to enhance complex protocols for enhanced features in small network scenarios. |
| Kim et al. [18] | CNN-LSTM | This study proposes a deep learning-based Haugh Unit (HU) prediction model to monitor egg quality in cold chain storage and transportation. The model uses non-destructively measured weight loss to determine egg freshness. | Leveraging weight loss data alone, the target transfer learning CNN-LSTM demonstrated a significant reduction in RMSE, decreasing from 6.62 to 2.02 compared to a random forest regressor. Furthermore, the application of data augmentation techniques resulted in a noteworthy decrease in the Mean Absolute Error (MAE) of HU prediction for the target model, dropping from 3.16 to 1.39. |
| Gao and Xu [19] | IoT | This paper presents a Wireless Sensor Network (WSN) for monitoring fruit cold storage using the ZigBee protocol. The system uses a Chip CC2430 for information processing and wireless node detection. | The system offers advantages such as automatic data acquisition, storage, and remote control for storage managers. |
| Feng et al. [20] | SVM | This study introduces a system that enhances transparency and trust by collecting and validating quality parameters. Employing K-means and SVM algorithms enables effective classification and prediction of quality loss. | The training set and test set proportions, falling within an allowable deviation range, stand at 88.89% and 87.17%, respectively. Utilizing the SVM model, the Root Mean Square Error (RMSE) for the training set is 0.1502, while for the test set, it is 0.1793. Notably, both the K-means and SVM models exhibit superior accuracy in their performance. |
| Pan et al. [21] | MultiLayer Perception Artificial Neural Network (MLPANN) | This paper proposes a hyperspectral imaging system to detect chill damage in peaches stored in cold storage. The system used an ANN model with eight optimal wavelengths. | Post-harvest storage-induced changes in factors such as firmness, soluble solids content, extractable juice, titratable acidity, and chlorophyll content were identified. ANNs pinpointed eight optimal wavelengths, achieving classification accuracies of 92.9%, 100%, and 95.8%. The MLPANN models demonstrated effective predictive performance for quality parameters. |
| Wagle et al. [22] | New compact CNN | This study introduced novel models, N1, N2, and N3, for plant leaf disease detection and assessed their classification accuracy and training efficiency in comparison to the AlexNet model. Notably, the proposed models exhibit a more compact design and require reduced training time compared to AlexNet. | The envisaged model effectively identifies diseased plant leaves in mobile phone-captured images. Its compact design allows deployment as a standalone mobile app, offering farmers an efficient solution with commendable classification accuracy. |
| Wagle et al. [23] | Three-layer CNN | This work proposed compact CNN models that reduced computational complexity to detect tomato plant leaf disease detection when compared to the ResNet-101 model | The suggested method demonstrates enhanced performance, attaining classification accuracies of 99.13%, 99.51%, and 99.40% for N1, N2, and N3 models, respectively. These models are characterized by their compact size, increased efficiency, and reduced time complexity, marking a substantial stride in the effective management of infected plants. |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.