Description:MACHINE LEARNING BASED USER – FRIENDLY PREDICTION OF PLANT DISEASES USING CLOUD OF THINGS
Field of Invention
The present invention is relating to the identification of plant disease using leaves for various species and provide medicinal information for diseases identified.
The Objectives of this Invention
This invention's major objective is to predict multiple plant diseases accurately and parallelly provides targeted treatment and minimizing the use of pesticides. The heterogenous uses of this system also include corresponding pesticide recommendations, reducing environmental impact. In the end we aim to provide a comprehensive, reliable and compact solution to protect crops and maximize the yield.
Background of the Invention
In (DE102019/201988A1), The current invention encompasses a number of methods for managing an agricultural framework, including an agricultural sunlight, a monitored agricultural system, and an agrarian management method. The disclosure also covers a technique for lighting up a plant cultivated in a farming environment for inspection (2620). In addition (US2020/0184153A1), The invention additionally pertains to an agricultural framework that includes an assortment of manufacturing lines for cultivating plants of a specific plant kind, in which a first analysing line in the diversity of analysing lines is set up to transport a first diversity of crops together a path via the farming system and employ a first development circumstance to the first diversity of crops in order to satisfy a first active ingredient variable for the first diversity of plant life.
Recent improvements in AI have greatly enhanced the accuracy and efficiency of Convolutional Neural Networks (CNN) for image identification. State-of-the-art CNN models now utilize advanced techniques such as attention mechanisms, neural architecture search, and transfer learning. These improvements have led to breakthroughs in various image identification tasks, including object detection, semantic segmentation, and image classification. With the continued advancement of AI and CNN technology, the potential for practical applications in fields such as healthcare, farming is rapidly expanding by (Udutalapally et al [2020], ArXiv, Vol. (abs/2005.06432), pp-1-16).
However, upon studying their work, it has become apparent that these systems are primarily focused on the diseases of a single crop. While some attempts have been made to include multiple crops, not all diseases have been adequately implemented by (Orchi et al [2022], Agriculture, 12(1):9).
To address this issue, this invention proposes the use of advanced Convolutional Neural Network (CNN) models to identify all diseases of multiple crops. Additionally, curing the diseases also plays an important role in crop yield. Therefore, a database for crop diseases and details of pesticides will be created, allowing for the recommendation of the best curing technique to enable maximum yield while minimizing unwanted soil pollution. By utilizing these advanced technologies and developing a comprehensive database of crop diseases and pesticide details, this invention aims to revolutionize the agriculture industry. It has the potential to not only improve crop yields but also reduce the environmental impact of pesticide use. This innovative approach to disease prevention and cure has the potential to greatly benefit farmers and consumers alike.
Summary of the Invention
The proposed invention will be helpful for the farmers to save time and capital. Firstly, the picture of plant leaf will be analyzed and results will be obtained. Using this system, we can access the critical information quickly and efficiently and provide the targeted treatment or take measures for early prevention of the disease.
Detailed Description of the Invention
The proposed invention for disease prediction supports live monitoring of multiple plant species, in this system, plant leaves are continually recorded and transmitted to a microcontroller where they are used to separate different kinds of plants based on earlier recognised plant location. Next, leaves are taken away from an image feed and transmitted to the cloud, in which we use an algorithm based on machine learning to determine regardless of whether or not a leaf is infected. We also maintain database of pesticides for diseases of different plant species, if any plant leaf is detected as diseased then accordingly a recommendation of pesticide is made for that plant species to cure or to stop the spread of disease. Detection of plant disease can be done by training algorithm by using CNN deep learning by the data collected. This collected data involves leaves which are diseased and non-diseased. Diseased leaves are again categorized or classified based on the type of disease. From the database of pesticides we have, we can able to recommend pesticide over internet to mobile application.
Data collection involves the collection of plant leaf images from various sources like Kaggle and plant nurseries, here we have collected datasets of potato, tomato, and bell pepper. There are further categories for each species. Potato has three labels, they are Potato Healthy, Potato Early Blight, and Potato Late Blight. Tomato has 10 labels, they are 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. Bell pepper has 2 labels, they are Bacterial Spot and Healthy. On by collecting data, cleaning operation is performed to remove unwanted blurry images. The total count of images after performing data cleaning is 20598. After the completion of data cleaning, we perform data partitioning. Here in this step, we partition our data into three types, which are training data, validation data, and test data in the ratio 7:1:2. After the partitioning it is found that the count of training data is 14440, the count of validation data is 2058, and the count of test data is 4140.
Training data is used to train the model, while validation data is used to validate batches of training data for every epoch(iteration), and test data is used to test the CNN-ML model before deploying the model. After partitioning the data, we perform data augmentation which includes resizing, rescaling, horizontal flip, and vertical flip. Data augmentation is done to generate multiple datasets from the existing dataset by performing the above operations, this is done to increase the accuracy of the model. After data augmentation, these images are converted into 3-D spatial data structures ranging from 0-255(RGB scale), now these are sent as input to the CNN model. CNN model mainly has 3 layers, which are convolution layer, activation layer, and pooling layer. The function of convolution layer is to perform feature extraction by applying filters. Dot product of 2-D matrix of image and filter is performed and the result of dot product is sent as an input to activation layer. In pooling layer dimensionality reduction is performed and this makes the tolerant towards various distortions, pooling is of 2 types, they are max pooling and average pooling, in our model we have performed max pooling. In this way we build our CNN model. Model will be deployed to cloud, from front end through web/mobile/microcontroller applications we send photo of leaf to backend where our model will process image and results are obtained.
The functions in user interface module includes sign in/sign up page and input/ output fields. This module allows only the authenticated users to login and then user can able to select the plant type and scan the leaf to identify any disease. It is used to deploy the machine learning trained module and image data feed from user applications and on by receiving, the data is sent for disease prediction. It is to clean the collected by evaluating it with manually collected data and setting a pipeline to clean the data and partition the data as validation, test and train data. Model will be trained using CNN deep learning.
The use case activity of the invention has the sequence of the actions in a process such as; Initially the user has to register to the application, after registration the user has to login into the application with respective login credentials. Once done with the login process, images are either captured by mobile or any other camera and images are sent as input feed. Now the extracted data is sent to the cloud server, the database in the cloud contains a large number of datasets which include both healthy and unhealthy leaves as well as the list of pesticides for the corresponding leaf disease detected. The ML CNN model is trained in such a way that based on the existing data set present in the database it predicts the presence of the disease. If any disease is predicted then the application notifies the user regarding the presence of the disease and corresponding cure pesticide is recommended to the user.
The proposed invention for disease prediction supports monitoring of multiple plant species, in this data of plant leaves is collected and through mobile application or web application these images can be transferred using internet then from the image feed leaves will be extracted and sent to cloud where we apply our machine learning algorithm to decide whether leaf is diseased or not. We also maintain database of pesticides for diseases of different plant species, if any plant leaf is detected as diseased then accordingly a recommendation of pesticide is made for that plant species to cure or to stop the spread of disease. Detection of plant disease can be done by training algorithm by using CNN deep learning by the data collected. This collected data involves leaves which are diseased and non-diseased. Diseased leaves are again categorized or classified based on the type of disease. From the database of pesticides we have, we can able to recommend pesticide over internet to mobile application.

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

Claim:
1. A system/method to the identification of plant disease using leaves for various species using machine learning and cloud of things technology, said system/method comprising the steps of:
a) The system starts up, and the home garden's stream of data is collected by the camera (1). The micro controller (2) is then provided with the information it has collected feed. It then requires a modem (3) for access to the web.
b) The developed system for deployment on a cloud-based platform accepts data feeds for analysis (4). Prediction is used in data stream analysis (5, 6). Using a user application, the user is alerted if it detects a problem with the data stream (picture) (7, 8).
c) The CNN method is used for incorporating a novel model that has been trained into the proposed innovation. Data collection (1), data pipelines (2), data cleaning (3), data partitioning (4), data augmentation (4), and training the CNN model (5), (6), are all included in this.

2. As mentioned in claim 1, the camera gathers data feed from the backyard garden and sends it to the microcontroller (IoT), which subsequently transmits the data feed via a connection for access to the internet.
3. According to claim 1, the developed system for deployment on a cloud-based service accepts data feeds for analysis. Data feed analysis is carried out using anticipation.
4. According to claim 1, the suggested CNN inventive paradigm examines the leaves and tells the user via an application that is used if there may be any issue with the data flow (picture).
5. According to claim 1, the data inputs are handled in many ways, including collection, pipelining, cleaning, augmentations, and finally predictions, in the process of training system parts.