Artificial Intelligence Based Food Quality Detection System

Abstract

In recent years, the modern food industry has been increasingly driven by consumer preferences for fresh, nutritious, and uncontaminated products, particularly in the case of fruits and milk. Ensuring food safety and quality is paramount, and to meet these demands, advanced technologies for quality detection have emerged. This research presents an innovative integrated system that harnesses the potential of artificial intelligence (AI), sensor technologies, and image analysis techniques to provide an efficient and reliable solution for food quality detection. AI algorithms can be trained to recognize patterns and anomalies in the sensory and image data collected from the food products. The integrated system addresses the limitations of conventional quality detection methods. Traditional methods often involve time-consuming and labor-intensive processes, making them less practical for real-time monitoring and assessment. In contrast, the AI-driven integrated system offers real-time and accurate assessments, significantly improving efficiency and reducing the time required for quality evaluation. This advancement is crucial in minimizing the risk of foodborne illnesses by promptly identifying potential hazards and ensuring compliance with safety standards. Sensor technologies further enhance the system's capabilities by collecting vital data on various parameters like temperature, moisture, pH levels, and chemical composition. These sensors provide crucial insights into the freshness and overall quality of the food products. Image analysis techniques, integrated into the system, contribute to assessing visual attributes such as color, texture, and visual defects. By merging AI, sensor technologies, and image analysis techniques into an integrated system, the food industry can achieve a comprehensive and efficient approach to quality detection. This innovation has the potential to revolutionize food safety practices, reduce food wastage, and ultimately meet the growing consumer demand for high-quality, safe, and nutritious food products. Fruits and dairy products are essential components of a balanced diet for most living beings, including humans, due to their nutritional content and health benefits as it contains necessary vitamins and minerals. Detecting fruit quality amidst complex backgrounds using an automated system holds paramount importance in the fruit industry. Among the critical factors in fruit grading, appearance plays a vital role, influencing market value and consumer preferences. However, there is a pressing need for an automated system to swiftly evaluate fruits, identify defects, and sort them based on their quality. Deep learning algorithms have revolutionized object detection, making a substantial impact on fruit quality detection. Notably, Mask R-CNN and YOLOv5 are two prominent object detection algorithms that have been thoroughly experimented with in this study. YOLOv5 particularly stands out, showcasing superior performance, especially in scenarios requiring real-time object detection. The proposed work focuses on developing a fruit identification and quality detection model based on the YOLOv5 object detection system. The dataset encompasses images of four distinct fruits—apple, banana, orange, and tomato—categorized based on their ripening index. The model operates in two fundamental phases: first, fruit identification is accomplished, followed by fruit quality detection. Mosaic augmentation, a technique applied during the training of phase 1, significantly enhances detection performance and robustness of the system. The model accurately classifies the fruit, and the predicted image is subsequently forwarded to phase 2 for precise fruit quality detection. Results demonstrate that the proposed method effectively identifies fruits and detects their quality using the validation dataset. Real-time performance tests have been conducted using sample inputs to showcase the system's efficiency. The fruit identification and quality detection model have been compared with several state-of-the-art detection methods, yielding highly encouraging results, solidifying its potential and effectiveness in the realm of fruit quality assessment. Additionally, most perishable fruit of the collected dataset tomato is further explored to evaluate shelf life of the fruit based on its ripeness index. A dedicated tomato quality detection system is introduced, focusing on analyzing specific attributes to ensure precise quality evaluation. It talks about an ensemble strategy to create a predictive system for tomato ripeness and shelf life, focusing on defects and color intensity. To ensure distinctiveness in the ensemble part of the proposed technique, a variety of expert base regressors, such as SVM, DT, RF, and GBM, are employed. Manual extraction of color, texture, and shape features is conducted, while the Inception V3 model automates feature extraction, subsequently reduced using the PCA dimensionality reduction method. The resulting ripeness regression models yield outputs categorized into three classes based on ripeness index and color magnitude, determining the tomatoes' shelf life as Store, Sell, or Discount. The stacking method is applied to consolidate these outcomes and produce the final prediction for the tomatoes' shelf life. The study reveals that incorporating different features, employing diverse pre-processing techniques, and utilizing skilled machine learning regressors significantly introduce variation into the ensemble approach, ultimately boosting accuracy compared to traditional machine learning models. Further, the research explores dairy products besides fruits to evaluate milk quality. The milk quality evaluation is a vital process that entails a comprehensive assessment of various attributes and components to ascertain the overall quality and safety of milk. Addressing the pressing concern of milk adulteration, which poses significant threats to the nutritional value of milk and the well-being of consumers, this research introduces a novel AI-powered Internet of Things (IoT) integrated multi-sensor system. The proposed system seamlessly integrates a range of sensors capable of real-time measurement, encompassing parameters such as pH, electrical conductivity (EC), temperature, gas properties, and Volatile Organic Compounds (VOC) parameters. This comprehensive approach extends to the measurement of crucial constituents within milk samples, including Fat, Protein, Solids Not Fat (SNF), Lactose, and Gravity values. To effectively identify specific adulterants—Urea, Starch, Sodium Bicarbonate, Maltodextrin, and Formaldehyde—a machine learning-based ensemble technique is employed for classification. This ensemble method surpasses conventional algorithms like RF, Light GBM, and Extra Trees Classifiers, achieving a notable accuracy rate in identifying adulterants present in the milk dataset. A significant aspect of this study is the creation of an IoT-based data acquisition device that seamlessly integrates with the sensor system, facilitating precise and efficient measurements. Additionally, SHAP (SHapley Additive exPlanations) analysis is utilized to elucidate the decision-making process of the ensemble model, thereby enhancing result interpretability. The system's ability to rapidly detect and categorize adulterants underscores its importance in mitigating the pervasive issue of compromised milk quality, thus ensuring consumer safety and upholding industry integrity. By integrating data from both fruit and milk quality detection modules, the system provides a comprehensive assessment of overall food quality. Leveraging the capabilities of AI and IoT, valuable insights are provided to producers, distributors, and consumers, empowering them to make informed decisions about food product freshness and safety. Extensive experiments using diverse datasets of fruits and milk samples validate the system's effectiveness, demonstrating high accuracy in detecting quality attributes and identifying potential risks associated with perishable food products. This pioneering research opens new avenues for leveraging AI and IoT technologies, paving the way for smarter, more efficient, and safer food supply chains, ultimately benefiting both producers and consumers alike. By ensuring consistent delivery of high-quality, untainted products, the integrated system aligns with consumer demands for wholesome and safe food options, while also contributing to elevated food safety standards and reduced economic losses due to spoilage and waste.