Description:A SYSTEM/METHOD TO DEFECT EXPOSURE IN VEGETABLES AND FRUITS USING MACHINE LEARNING ALGORITHMS
Field of Invention
The innovation relates to the use of machine learning algorithms to identify defect exposure in vegetables and fruits.
The Objectives of this Invention
The main objective of the innovation is being proposed here is to monitor and detect diseases in fruits and vegetables in their early stages to farmers for a healthy yield. Our System can deliver the most accurate results in a very short time while consuming minimal computational resources.
Background of the Invention
In recent years, farmers want to identifying the leaf diseases prediction in a effective manner. First, type of technique has been introduced in (US2019/10839503B2), This process uses an image of the crops or vegetables to generate image analysis findings, analyzing hyperspectral and/or Near Infrared (NIR) lighting reflected from the plant-based or fruit to generate reflection analysis outcomes, and formulating a minimum of one value corresponding to a minimum of one property of the plant-based or fruit determined by both the image analysis outcomes and contemplation evaluation outcomes are all possible steps in a system and procedure for effectively establishing the features of an agricultural product or fruits and vegetables. Another method, (US2019/0203267A1), Within this are given techniques and tools for distinguishing between harmful and non-harmful organisms in specimens from food and the environment. Existing difficulties in recognizing alimentary pathogenic microorganisms, such as Salmonella, Campylobacter, Listeria, and Escherichia, promptly and efficiently are resolved by the disclosure. Procedures for discriminating between a temporary and a resident pathogen, for connecting the presence of non-pathogenic with pathogenic microorganisms, and for distinguishing between live and dead microbiological by genotyping are also provided in the disclosure. In (US2020/11531317B2), This intelligently guided mechanism and technique for evaluating fruit and vegetables use a conveyor system to transport the goods, a variety of acquisition and processing hardware and software, water and air jets for eliminating and overseeing the orientation and positioning of the goods, and networking equipment and programs, all of which work in concert with one another to maximise efficiency and achieve the goals of high throughput with minimal loss of product quality. Unlike any other system on the market, the second-generation strawberry decaying system (AVID2) makes use of a convolutional neural network, also known as (AVIDnet) to make an unfair treatment network decision—namely, whether or not a strawberry is to be cut or rejected—and to calculate a multi-point cutline bending along which strawberries can be rapidly robotically cut. In (RU2017/2750085C2), The technique and mechanism of the new invention are used to conduct optically analysis of produce. At least one color camera (4) sensitive to infrared emission is used for creating visuals. At the same time, several light sources are set up to preferentially supply the emission of light in different spectral categories to each item following the established illumination sequences. The camera's exposition is timed to coincide with the row of lights, capturing photos throughout a wide range of wavelengths. This includes both visible and infrared light. The result is efficient, much more straightforward, and more affordable to learn, operate, and maintain techniques and instruments for optical characterizing crops and fruits through image production in different spectral bands.
The (Anupriya et al [2023], 2023 International Conference on Computer Communication and Informatics (ICCCI), Coimbatore, India, 2023, pp. 1-6), In the field of agriculture, it is tough to detect crop illnesses. Inaccurate detection can devastate crop yields and the value of the harvested goods on the market. This points to developing novel methods for identifying crop illnesses and tracking their spread. As a result, we suggest a new model in which convolution networks are used to categorize diseased leaves. The recommended method can differentiate infected leaves from healthy ones, allowing it to detect thirteen distinct types of crop illnesses. The primary motivation for introducing the model for detecting and predicting plant diseases is to serve those in the farming and agro-based industries. Improving crop yields aids in monitoring crop diseases, allowing operators to take preventative actions.
Summary of the Invention
Since India is primarily an agricultural nation, its farmers have a broad variety of plant options from which to choose. Our algorithm can detect the various defects in different fruits and vegetables, allowing for more precise cultivation procedures. Many small farmers in the countryside would benefit from our ability to save time and money by spotting flaws in their crops before planting. Our model's accuracy improves from 80 to 100 percent as epochs increase to 20. The initiative's primary goal is to ease the workload of humans, specifically farmers.
Detailed Description of the Invention
The two-step procedure of the described innovation includes the following steps:
We're taking our analyses to new heights. Its purpose is to verify that the trained model can be implemented. Following one's infinite resources and without any time limit, any invention or application model is possible. Time and product availability are crucial considerations. For a model to be workable, it must stay between time and money constraints.
Three distinct feasibilities exist Feasibility on multiple fronts, including the technical, operational, and financial levels.
Any model's viability can be gauged by how readily users can access it in a format that suits their needs. For it to be feasible, it must have a well-designed user interface and website. The World Wide Web is the most ubiquitous technology in today's developed model. The system must remain free of any technological problems and flaws at all times, including during periods of high demand. This feasibility is met if the product can be used by anyone with little or no training and it doesn't require an expert to explain how it works. In this section, we assess the level of product knowledge among consumers. Users must know how to apply the latest technological advances to their situations.
To determine whether or not a model is economically viable, we need to look at several factors, including Maintenance expenses, Long-term profits, and Cost-benefit analysis. These days, the information superhighway, often known as the internet, is a must-have for quick and easy access. Every business that needs it should be able to quickly and inexpensively acquire it. It must be optional or add extra work.
While there are many different kinds of learning, each with its own algorithms, supervised learning is the most well-known and practically viable option. Since the input data and desired outcome are known in advance, this sort of learning is easier. In this learning fashion, the model is initially taught employing the provided data, possibly allocating different quantities of data to training and testing. The results would be exceedingly accurate because they are based on a real-world scenario, the gold standard of learning. All that has to be done using the available data is to verify the input and locate the corresponding output. Input and output data would be given to the model, allowing it to learn rapidly and effectively to produce accurate results. This model fits the x and y categories. This learning serves various functions and has many real-world applications, including but not limited to any linear model, forecasting, Image classification, and many others.
The initial step of image pre-processing is to enhance the quality of the delivered photos so that they may be more easily analyzed later on. This is completed before we send our photographs to train and validate the model. It'll be helpful to in getting a high-resolution and quality photograph to work with. The features can also be improved with the use of image processing. The data is transformed so that it may be transmitted through a network. It also aids in enhancing the prediction model's precision and efficiency. Several methods of image processing are described here. Image processing techniques include cropping and filtering, intensity adjustment, histogram equalization, noise removal, and edge detection.
Cropping an image involves erasing irrelevant or distorted areas that would otherwise influence the prediction model's output. This is the first stage of the image processing pipeline. Using Tensor Flow, we can train the model with this pre-processing method and its associated phases. It helps regulate the level of brightness present at each image's pixel. The photos will be managed using equalization algorithms and compared to satellite-observed images accessible from various sources. They facilitate the pre-processing of pictures and are categorized into numerous subsets handled locally and globally.
We begin processing every image by identifying its edges and then classifying it. Images' dimensions must be established. Here, the image function is crucial for describing the picture edges and detecting them efficiently for defining partial derivatives, all of which contribute to developing the image technique. CNN is preferred over Neural networks since the former are weaker and unreliable. Each layer in a convolutional neural network (CNN) works separately to analyze and learn picture data and provide output predictions. It will function and think like the human brain. Still, it will also incorporate characteristics of other intelligent species, some of which have superior accuracy and precision in their abilities and, hence, are closer to the truth. While other neural networks extract information from input data, CNN works directly with imagery. Defect exposure in vegetables and fruits using machine learning algorithms can greatly benefit the agriculture and food industries. Let's create a simple use case to illustrate how this can work:

A tomato processing facility wants to automate the quality inspection process to detect defects such as bruises and spots in tomatoes. They aim to reduce manual labor and improve the consistency of tomato quality assessment.
Gather a dataset of tomato images, including both healthy and defective tomatoes. These images can be captured using cameras in the production line. Label each image as "Healthy" or "Defective" based on visual inspection. Resize and normalize the images to a common resolution. Split the dataset into training and test sets.
Use Convolutional Neural Networks (CNNs) to extract features from the tomato images. CNNs are well-suited for image classification tasks. For simplicity, let's use a pre-trained CNN model (e.g., VGG16, ResNet) and fine-tune it for our specific task. This can save time and computational resources. Train the fine-tuned CNN model on the training dataset, adjusting the final output layer to have two classes: "Healthy" and "Defective." Use techniques like data augmentation to increase the diversity of training samples and reduce overfitting. Assess the model's performance on the test dataset using metrics like accuracy, precision, recall, and F1-score. Perform visual inspections of misclassified images to understand potential areas for improvement. Integrate the trained model into the tomato processing line. Install cameras at the inspection point where tomatoes are sorted. In real-time, as tomatoes pass through the camera, capture images and use the model to classify them as healthy or defective. Implement an automated sorting system that diverts defective tomatoes to a separate conveyor belt for further processing or disposal. Healthy tomatoes can continue along the main production line. Continuously monitor the system's performance in the production environment. Retrain the model if necessary with new data to adapt to changing conditions. With this automated tomato quality inspection system in place, the processing facility can significantly reduce manual labor, improve accuracy in defect detection, and ensure consistent tomato quality. This not only enhances the overall quality of their products but also increases operational efficiency.
This use case demonstrates how machine learning algorithms, specifically CNNs, can be applied to detect defects in vegetables and fruits, leading to practical improvements in the agricultural and food processing industries.

4 Claims & 1 Figure
Brief description of Drawing
In the figure which are illustrate exemplary embodiments of the invention.
Figure 1, The Process of Proposed Invention , Claims:The scope of the invention is defined by the following claims:

Claim:
1. A system/method to detect and exposure the fruits or vegetables using machine learning algorithms, said system/method comprising the steps of:
a) The system starts with datasets collection from various cameras (1), from that problem statements has been framed (2).
b) The system is incorporated with identifications of required data (2), preprocessing steps (3), to identify some of the important predictable images (4), the algorithm selection process (5), the image is matched and accuracy metric was compared in evaluation test(6), then finally the classification model is triggered and produce the predicted output (7).
2. As mentioned in claim 1, the invented system starts with various videos and image dataset uploading to start the process.
3. According to claim 1, the preprocessing will initiate to remove the noisy data from the dataset and it will trigger feature extraction process of CNN algorithms to split the data into training and testing part.
4. According to claim 1, now, the proposed invention will start from CNN functions, then this will be matched with captured figure and detects the defected fruits or vegetables and type of fruits or vegetables using CNN architecture based machine learning algorithms.