A method for determining the freshness of aquatic products

Abstract

The invention relates to a computer-based method developed for determining the freshness of aquatic products (such as fish) through image processing and artificial intelligence-based algorithms.

Description

[0001] A METHOD FOR DETERMINING THE FRESHNESS OF AQUATIC PRODUCTS

[0002] Technical Field

[0003] The invention relates to a computer-based method developed for determining the freshness of aquatic products (such as fish) through image processing and artificial intelligence-based algorithms.

[0004] Prior Art

[0005] In the art, various traditional methods are used to determine the freshness of aquatic products. These methods generally include sensory evaluation, chemical and biological analyses, and in some cases, imaging techniques.

[0006] Sensory Evaluation: The properties of the product such as visual, odor, taste, and texture are evaluated by experts or trained personnel. This method is based on the experience of the experts and is subjective. At the same time, it does not always give fast results and it can be difficult to evaluate the freshness of the products accurately.

[0007] Chemical and Biological Analyses: Samples are taken from aquatic products and chemical and biological analyses are carried out in the laboratory. These analyses include the measurement of various components that provide information about the freshness status of the products. The analyses are generally time-consuming and costly. In addition, laboratory infrastructure and expert personnel are required for the analysis.

[0008] Imaging and Color Analysis: Special camera systems are used to examine the colors and visual features of the products. Color changes provide information about the freshness status of the products. Such systems are generally expensive and are usually found in large-scale enterprises. Also, imaging techniques most of the time remain limited to color analysis. CN116087195A discloses a fish freshness evaluation method and system.

[0009] WO20 18093142 Al discloses a method for inspecting fish freshness and a device for this method.

[0010] CN106556602A discloses a fish freshness determination method and a device related thereto.

[0011] When existing studies in the art are examined, a need has arisen for developing the invention-related method that enables determining the freshness of aquatic products (such as fish) through image processing and artificial intelligence-based algorithms.

[0012] Objectives of the Invention

[0013] The object of this invention is to develop a computer-based method that enables determining the freshness of aquatic products (such as fish) through image processing and artificial intelligence-based algorithms.

[0014] Another object of this invention is to develop a computer-based method that allows evaluating the freshness of aquatic products within seconds.

[0015] Another object of this invention is to develop a computer-based method that eliminates dependence on sensory evaluations and makes freshness analysis objective.

[0016] Detailed Description of the Invention

[0017] The invention relates to a computer-based method developed to determine the freshness of aquatic products through image processing and artificial intelligencebased algorithms, it comprises,

[0018] Taking a photo with a mobile phone camera so that the fish completely fills the screen,

[0019] Detecting whether the fish fits on the screen using the OpenCV library, Applying Gaussian Blur to the photo to reduce image noise, Determining the edges of the fish in the photo using the Canny Edge Detection method,

[0020] Determining the region of interest (ROI) value of the fish and checking whether these ROI values are within the screen boundaries,

[0021] Calculating the average brightness value of the image using the histogram analysis method,

[0022] Detecting the fish species using the Support Vector Machines (SVM) algorithm,

[0023] Performing image processing again using the OpenCV library,

[0024] Applying filtering with the Non-local Means algorithm to reduce image noise,

[0025] Determining the edges of the fish using the Canny Edge Detection method, Determining the main outlines of the fish by contour analysis and defining it as the region of interest,

[0026] Increasing image contrast by histogram equalization and making the fish’s skin area more distinct,

[0027] Converting from the RGB color space to the Lab* color space using the OpenCV library,

[0028] Detecting whether the fish is fresh or spoiled using the Support Vector Machines algorithm with a labeled dataset consisting of Lab* values of fresh and spoiled fish.

[0029] The stages of the inventive method are explained in detail below. To avoid ambiguity in the explanation of the invention, the inventive method is described only based on fish. However, the inventive method can also be used for other aquatic products instead of fish.

[0030] In the inventive method, first, the user opens the phone camera and takes a photo of the fish so that the fish completely fills the screen. After the photo is taken, it is detected in real time using OpenCV whether the fish fully fits on the screen. Gaussian Blur is applied to reduce image noise. The edges of the fish are determined by the Canny Edge Detection method. The region of interest (ROI) of the fish is determined, and it is checked whether this ROI is within the screen boundaries. Instructions such as "Move away" or "Move closer" are given to the user when the fish does not fully fit on the screen.

[0031] Then, the average brightness value of the image is calculated using histogram analysis and brightness detection methods. The proportion of pixels above a certain brightness threshold is analyzed. Instructions such as "The light is too bright" or "Reduce the light" are given to the user.

[0032] Support Vector Machines (SVM) and Random Forest models are trained on a large dataset defining fish species. Then, fish species-specific features are extracted from the image. The species of the fish is identified and notified to the user. After the fish species determination, freshness criteria appropriate to the species are determined by the method. This is because the skin color of each fish changes in a unique way, and accordingly, the freshness criteria also vary.

[0033] After the fish species is determined, image processing is performed using the OpenCV library. To reduce image noise, filtering is applied with the Non-local Means algorithm. The edges of the fish are determined by the Canny Edge Detection method. The main outlines of the fish are determined by contour analysis and defined as the ROI. The image contrast is increased by histogram equalization, and the fish’s skin part is made more distinct.

[0034] In the color analysis stage, in the Lab* color space,

[0035] L (Lightness): varies between 0 (black) and 100 (white). a (Green-Red Axis): Negative values represent green tones, positive values represent red tones. b (Blue- Yellow Axis): Negative values represent blue tones, positive values represent yellow tones. Using OpenCV, a conversion is made from the RGB color space to the Lab* color space. Pixel values within the determined ROI are analyzed, and the average L*, a*, and b* values are calculated.

[0036] A histogram is created to analyze the color distribution within the ROI.

[0037] A labeled dataset consisting of Lab* values of fresh and spoiled fish is used. For each fish, the average L*, a*, and b* values are extracted as features. The model is trained using the training dataset. The accuracy of the model is tested with cross- validation. The Lab* values of the live-captured image are analyzed by the model. Based on the Lab* values, the model determines whether the fish is fresh or spoiled and reports the result to the user.

[0038] In the inventive method, analysis results, user information, and captured photos are securely stored using MySQL or PostgreSQL.

[0039] The advantages of the Support Vector Machines (SVM) and Random Forest algorithms used in the invention-related method over other algorithms in the art are also specified below.

[0040] Data usage efficiency:

[0041] SVM and Random Forest perform better than CNN, especially when working with limited amounts of data. While CNN typically requires a large amount of training data, SVM and Random Forest can achieve effective results with less data. This is particularly beneficial for data-limited situations such as aquatic product freshness.

[0042] Computational resources:

[0043] SVM and Random Forest generally require fewer computational resources. Since CNN models are deep learning structures with many parameters, their training processes usually take longer and require more computational power. SVM and Random Forest, being simpler in structure, train faster and consume fewer resources. Model transparency and interpretability:

[0044] SVM and Random Forest are more transparent in terms of understanding how the model makes decisions. CNN is often referred to as a "black box" because its internal workings and decision-making processes can be difficult to interpret. Especially in fields such as food safety and quality control, having a clear and understandable decision-making process is preferred.

[0045] Feature importance and selection:

[0046] Random Forest can evaluate the contribution of each feature to the classification result, which is used to determine which features are more important. SVM identifies which data points are critical for classification by using support vectors. These features are important for understanding which image characteristics the model is based on.

[0047] Innovative Aspect of the Invention and Difference from Other Inventions

[0048] The inventive method can make effective predictions with a limited amount of aquatic product data. It offers fast data processing and analysis suitable for realtime applications. It provides the necessary transparency and reliability for freshness control throughout production and the supply chain.

[0049] The inventive method eliminates dependence on sensory evaluations and makes freshness analysis objective. Image analysis always provides results based on the same criteria and ensures consistency.

[0050] Objective evaluation ensures consistent assessment of the quality of aquatic products. This increases product safety and consumer satisfaction, while also preventing problems that may arise due to incorrect results.

[0051] The invention significantly reduces costs by eliminating the need for laboratory analyses or expensive equipment. Since it is a software -based solution, operating and maintenance costs are also low. Low cost makes this technology more accessible for small and medium-sized enterprises. This allows more businesses to monitor the freshness of their products, which in turn increases quality across the industry.

[0052] The software can continuously monitor the freshness status of aquatic products and store this data, enabling trend analysis. This provides valuable insights in product management and quality control processes. Real-time data allows tracking changes in product freshness and process optimization. Additionally, it continuously improves production and storage techniques, which increases long-term operational efficiency.

[0053] A simple and intuitive user interface enables even users without technical knowledge to fully benefit from the system. Its user-friendliness allows reaching a wider audience.

[0054] Easy accessibility and use enable the rapid adoption of the technology throughout the industry. This, in turn, popularizes quality control and freshness assessment.

[0055] In alternative applications, a similar freshness assessment system can be created using different machine learning and image analysis algorithms. Alternative solutions can be developed by using techniques such as infrared or thermal imaging.

[0056] The freshness status of aquatic products can be determined with sensors that detect chemical properties. Similar-purpose devices can be created to determine freshness status using microbiological data.

[0057] Regarding the use of the inventive method in other fields, the method can be used for freshness analysis of other food products such as meat, dairy products, fruits, and vegetables. It can be used for quality control during the storage and distribution of food products. It can be used to determine the quality and health status of biological samples and cultures. It can be used in hospitals or laboratories to monitor the condition of medical samples. It can analyze the color of water to monitor water quality in lakes and rivers. It can be used to preserve biological diversity and monitor ecosystem health in seas and lakes. Fish processing facilities and distribution centers can use the invention-related method to quickly determine the freshness status of products. This is especially important in quality control processes and for optimizing the shelf life of products.

[0058] Supermarkets, restaurants, and fish markets can use this application to evaluate product shelf life and offer fresh products to customers. Freshness assessment provides higher customer satisfaction and sales efficiency. Aquatic product breeders and farms can use it to assess post-harvest freshness of products. This ensures that products are fresh before being presented to the market.

[0059] Logistics and storage companies can use it to monitor freshness during the transportation and storage of products. This helps ensure that products are preserved under appropriate conditions before reaching consumers.

[0060] In summary, the inventive method can be applied across various sectors in the food industry to improve quality control, customer satisfaction, and operational efficiency.

Claims

CLAIMS1. A computer-based method developed to determine the freshness of aquatic products through image processing and artificial intelligence-based algorithms, characterized in that it comprises,Taking a photo with a mobile phone camera so that the fish completely fills the screen,Detecting whether the fish fits on the screen using the OpenCV library,Applying Gaussian Blur to the photo to reduce image noise, Determining the edges of the fish in the photo using the Canny Edge Detection method,Determining the region of interest (ROI) value of the fish and checking whether these ROI values are within the screen boundaries, Calculating the average brightness value of the image using the histogram analysis method,Detecting the fish species using the Support Vector Machines (SVM) algorithm,Performing image processing again using the OpenCV library, Applying filtering with the Non-local Means algorithm to reduce image noise,Determining the edges of the fish using the Canny Edge Detection method,Determining the main outlines of the fish by contour analysis and defining it as the region of interest,Increasing image contrast by histogram equalization and making the fish’s skin area more distinct,Converting from the RGB color space to the Lab* color space using the OpenCV library,Detecting whether the fish is fresh or spoiled using the Support Vector Machines algorithm with a labeled dataset consisting of Lab* values of fresh and spoiled fish.

2. A method according to claim 1, characterized in that the random forest method is used instead of the support vector machines method.