SYSTEM/METHOD TO IDENTIFY DIFFERENT MEDICINAL PLANTS/RAW MATERIALS

Abstract

The proposed invention is a user-friendly system that makes use of machine learning, more specifically convolutional neural networks, to accurately identify medicinal plants from photographs by making use of the botanical characteristics of the plants under consideration. The primary objectives consist of the curation of a comprehensive dataset, the development of an interface that is simple to use, and the provision of users with substantial information on the medicinal characteristics and traditional applications of the plants that have been found. In addition to assisting with accurate plant identification through the integration of traditional herbal knowledge and cutting-edge technology, the system offers recommendations that are dependent on the location of the plant, which is particularly useful in urban settings. Priority is given to the implementation of measures to preserve privacy, the incorporation of feedback, and continual model optimization. The ultimate goal is to provide users with a tool that adheres to ethical, cultural, and legal standards, encourages the use of plants in a sustainable manner, and contributes to the preservation of biodiversity. Through the facilitation of sustainable plant use, the assistance in the conservation of biodiversity, and the provision of users with a tool that adheres to ethical, cultural, and legal standards, the ultimate objective is to promote a harmonious fusion of traditional knowledge and technological innovations in the field of medicinal plants. 5 Claims & 1 Figure

Specification

Description:SYSTEM/METHOD TO IDENTIFY DIFFERENT MEDICINAL PLANTS/RAW MATERIALS
Field of the Invention
The present invention is relating to a system and method for identifying different medicinal plants/raw materials through Image Processing using ML Algorithms.
Objectives of the Invention
With the help of machine learning, this idea seeks to develop a user-friendly system that can correctly identify medicinal plants from photos and provide details on their traditional usage and health advantages. By fusing traditional knowledge with technology, the intention is to make it simple for consumers to find these plants, especially in cities. The main goals of privacy protection and continuous enhancements are to support sustainable plant usage and biodiversity preservation.
Background of the Invention
The patent application number (US2020/10856807B2) describes a system and method for assessing components of one or more objects at the same time based on the analysis of picture data and item component data, as well as interpreting such data for the user. Users can use a computing device to input item data, and the system can recognize or identify items through image recognition, optical character recognition, voice recognition, typed query, barcode scan, or any combination of these methods. The system then analyzes each item, returning it to the user's computing device via visual display, audible communication, or a combination of both visual display and audible communication. The system and method utilize image databases, image recognition services, application program interfaces (APIs), and machine learning to execute the analysis of items, recognize or identify their images, and analyze their components. The system and method aim to provide the user with item analysis by processing the data associated with those items.
Moreover, according to the (US20230111370A1), different aspects of the current disclosure refer to a technique for counting the number of colonies. On a colony enumeration device, we use a pre-trained deep learning model to identify colony-forming units of microorganisms in a composite image. This includes the identification of the microorganisms. The method can send a plurality of colony-forming units' identification characteristics to an interaction component. This is done so that the interaction component can project at least some of the plurality of identification characteristics onto the combined image.
(US2022/11372394) As stated in Methods and systems for a monitoring system for data collection in an industrial environment including a data collector communicatively coupled to a plurality of input channels connected to data collection points operationally coupled to an industrial chemical process; a data acquisition circuit structured to interpret a plurality of detection values from the collected data, each of the plurality of detection values corresponding to at least one of the plurality of input channels; and an expert system detection circuit structured to detect a process indicator in response to the plurality of detection values, and to initiate a self-organizing data collection response as a result of the process indicator detection.

Additionally, the patent number (US2021/0209705A1) protects the system and method for managing and maintaining a closed-loop agricultural-origin-product manufacturing supply chain network. An example of a method is the collection of agricultural data from multiple sources that pertain to multiple growing plots of crops; the collection of environmental data that pertains to the multiple growing plots; the collection of operational data that pertains to the intended utilization of the crops at a manufacturing facility; the identification of a specific growing plot; the correlation between agricultural data related to the specific growing plot, environmental data related to the specific growing plot, and operational data related to the intended utilization of crops from the specific growing plot; and the identification of a particular growing plot. We examine the correlated data to produce agricultural action recommendations for the specific growing plot and operational action suggestions for the manufacturing plant.
The method and apparatus for predicting the properties of a target object rely on a search manager to analyze the parameters of multiple databases. These databases include an electrical, electromagnetic, and acoustic spectral database (ESD), a micro-body assemblage database (MAD), and a database of image data. These databases store data objects that contain identifying features, source information, information on site properties and context, as well as time- and frequency-varying data. The method uses content-based image retrieval, multivariate statistical analysis, and principal component analysis to give two-dimensional information about three-dimensional objects. For example, it prefers image segmentation using a tree of shapes. Furthermore, the method is able to predict additional object properties using k-means clustering and other related methods. For instance, we can forecast, locate, and qualify events related to criminal activity, fraud, intrusion, and fire, as well as residual objects. As an example, we can check the properties of these leftover objects using black body radiation and micro-body databases that have charcoal assemblages, as explained in (US2014/8762379B2).

This server also offers a method and structure for provisioning applications, as stated in (US2019/10504020B2). This server provides a service known as application provisioning. A client user creates a schema that defines an application. The application interacts with peripherals attached to the client and also receives data input from sensors coupled to the peripherals. The sensors send the collected data to the server, where various methods, including neural networks, process it. The application includes a workflow that defines a finite-state machine. The response to sensor input, at least partially, determines how this machine traverses' states. It is possible for the server to offer dynamic reallocation of compute resources in order to satisfy the demand for classifier training task requests; the utilization of jurisdictional certificates in order to define data usage and sharing; and the fusion of data.
Summary of the Invention
In order to save both time and money, the farmers will benefit from the idea that has been offered. An examination of the photograph of the plant leaf will be carried out first, and the results will be acquired. By utilizing this system, we are able to have access to the essential information in a timely and effective manner, giving us the ability to administer focused therapy or to take preventative actions against the disease at an earlier stage.
Image processing and machine learning are two of the methods that will be utilized in this research in order to accomplish the goal of developing a practical instrument for the identification of medicinal plants. The development of a mobile application that will facilitate this process in an effective manner is the primary focus of this endeavor. Additionally, MongoDB is incorporated into the project in order to facilitate efficient data management. This research has the potential to have an impact not only on the field of herbal medicine but also on the field of biodiversity conservation.
Detailed Description of the Invention
The suggested invention for disease prediction allows for live monitoring of several plant species. Within this system, plant leaves are continuously recorded and transferred to a microcontroller, where they are utilized to differentiate between different types of plants based on the location of the plant that was previously identified. Next, leaves are removed from an image stream and sent to the cloud, where we apply an algorithm that is based on machine learning to assess whether or not a leaf is contaminated. This process is repeated until the leaves are no longer infected. Furthermore, we have a database of pesticides that are effective against illnesses that affect various plant species. In the event that a leaf of a plant is identified as being infected, a pesticide recommendation is produced for that particular plant species in order to either treat the disease or prevent its further spread. Through the use of CNN deep learning and the data that has been acquired, it is possible to detect plant diseases through the training of an algorithm. Leaves that are diseased and leaves that are not diseased are included in this data collection. Leaves that have been affected by disease are once more categorized or classified according to the type of disease. The pesticides that we have in our database allow us to make recommendations for pesticides that may be accessed via the internet or a mobile application.
The process of collecting data entails gathering photos of plant leaves from a variety of sources, such as Kaggle and plant nurseries. In this particular instance, we have gathered datasets of potato, tomato, and bell pepper. Every species is further subdivided into several categories. There are three labels: Potato Healthy, Potato Early Blight, and Potato Late Blight. Potatoes include all three of these labels. There are ten different diagnoses that can be applied to tomatoes. These diagnoses include Tomato Mosaic Virus, Target Spot, Bacterial Spot, Yellow Leaf Curl Virus, Late Blight, Leaf Mold, Early Blight, Spider Mites Two-Spotted Spider Mite, Tomato Healthy, and Septoria Leaf Spot disease. There are two labels on bell peppers: the Bacterial Spot mark and the Healthy label. A cleaning process is carried out once the data has been collected in order to get rid of any undesired fuzzy photos. Following the completion of the data cleaning process, the total number of photos is 20598. The process of data splitting is carried out once the data cleansing process has been finished. Within this stage, we divide our data into three categories: training data, validation data, and test data, with the proportions of each being seven, one, and two respectively. It was discovered that the count of training data is 14440, the count of validation data is 2058, and the count of test data is 4140. This was discovered after the partitioning process was completed.
Validation data is used to validate batches of training data for each epoch (iteration), and test data is used to test the CNN-ML model before it is deployed. Training data is used to train the model, while validation data is used to validate batches of training data. Following the segmentation of the data, we begin the process of data augmentation, which comprises resizing, scaling, flipping the data horizontally, and flipping it vertically. It is possible to build multiple datasets from the existing dataset through the process of data augmentation, which involves performing the activities described above. This is done in order to improve the accuracy of the model. As a result of the data augmentation process, these images are transformed into three-dimensional spatial data structures with a range of 0-255 on the RGB scale. These data structures are then used as input for the CNN model. The convolution layer, the activation layer, and the pooling layer are the three layers that make up the majority of the CNN model. Feature extraction is the task that the convolution layer is responsible for carrying out through the application of filters. After performing a dot product on the two-dimensional matrix of the image and the filter, the result of the dot product is then introduced into the activation layer as an input. In the process of pooling, dimensionality reduction is carried out, which results in the layer becoming more tolerant of a variety of distortions. There are two types of pooling: max pooling and average pooling. In our model, we have utilized max pooling. Through this method, we construct our CNN model. The model will be distributed to the cloud, and from the front end, we will send a photo of the leaf to the back end, where our model will process the image and receive the results. This will be done using online, mobile, and microcontroller applications.
A sign-in/sign-up page, as well as input/output fields, are among the functionalities that are included in the user interface module. Through the usage of this module, only users who have been authenticated are able to log in. Once logged in, users are able to select the type of plant and scan the leaf in order to identify any diseases. For the purpose of disease prediction, it is utilized to install the machine learning trained module as well as picture data feed from user apps. Additionally, the data sent for disease prediction is received. In order to clean the acquired data, it is necessary to evaluate it using data that was collected manually, install a pipeline that will clean the data, and divide the data into three categories: validation, test, and train data. CNN deep learning will be utilized in the training of the model.
The use case activity of the invention includes the sequence of actions that occur in a process, such as: Initially, the user is required to register to the application, and then, following registration, the user is required to log in to the application using the appropriate login credentials. As soon as the process of logging in is complete, photographs are taken using either a mobile device or any other camera, and then these images are transmitted as an input stream. At this point, the data that has been extracted is transferred to the server that is located in the cloud. The database that is located in the cloud contains a huge number of datasets that include both healthy and unhealthy leaves, as well as a list of pesticides that are used to treat the leaf disease that has been identified. The machine learning convolutional neural network (CNN) model is trained in such a way that it can predict the presence of the disease based on the data set that is already present in the database. If any disease is predicted, the application will notify the user about the presence of the disease, and the application will also recommend a pesticide that is effective in treating the disease according to the prediction.
The proposed invention for disease prediction allows for the monitoring of multiple plant species. In this process, data of plant leaves is collected, and images of the leaves can be transferred using the internet through a mobile application or a web application. The leaves will then be extracted from the image feed and sent to the cloud, where our machine learning algorithm will determine whether or not the leaf is diseased. Furthermore, we have a database of pesticides that are effective against illnesses that affect various plant species. In the event that a leaf of a plant is identified as being infected, a pesticide recommendation is produced for that particular plant species in order to either treat the disease or prevent its further spread. Through the use of CNN deep learning and the data that has been acquired, it is possible to detect plant diseases through the training of an algorithm. Leaves that are diseased and leaves that are not diseased are included in this data collection. Leaves that have been affected by disease are once more categorized or classified according to the type of disease. The pesticides that we have in our database allow us to make recommendations for pesticides that may be accessed via the internet or a mobile application.

5 Claims & 1 Figure
Brief description of Drawing
In the figure which are illustrate exemplary embodiments of the invention.
Figure 1. Flowchart showing the steps involved in plant identification. , Claims:The scope of the invention is defined by the following claims:

Claim:
1. A system/method to the identification of different medicinal plants/raw materials through image processing using machine learning algorithms, said system/method comprising the steps of:
a) The system starts with the proposed system environment (1), to collect or gathered data from various images (2), then apply the preprocessing steps to remove the images (3, 4).
b) The developed system will processed the inputs as 70:30 (5), and evaluate the performance with various metrics (6). The proposed model will analyzed and predict the results (7, 8, 9).

2. According to claim 1, An innovative system that uses Convolutional Neural Networks (CNNs) to accurately identify medicinal plants based on botanical features sets itself apart by combining deep learning techniques with an integrated approach for improved accuracy.
3. As per claim 1, Providing users with extensive details on recognized plants, such as characteristics, historical applications, and health advantages, is a unique feature that distinguishes it as an integrated knowledge distribution system that goes beyond traditional image classification.
4. According to claim 1, A unique right for giving users information about locations where medicinal plants are available; especially helpful for users in areas where access to these plants is restricted, this information promotes accessibility and user empowerment.
5. According to claim 1, the proposed system designed specifically for the identification of medicinal plants, offering a special set of frameworks and tools for image processing and machine learning for plant identification.

Documents