Plant Disease Detection by means of K-Means Clustering and HSV Algorithm R. Anirudh Reddy, Assistant Professor, Department of ECE, B V Raju Institute of Technology, Narsapur, Medak Dt, Telangana. anirudhreddy.r@bvrit.ac.in K. Sai Prasanna, Assistant Professor, Department of ECE, B V Raju Institute of Technology, Narsapur, Medak Dt, Telangana. Saiprasanna.k@bvrit.ac.in B. Mani Teja, Department of ECE, B V Raju Institute of Technology, Narsapur, Medak Dt, Telangana. 19211a0430@bvrit.ac.in N. Deepak Reddy, Department of ECE, B V Raju Institute of Technology, Narsapur, Medak Dt, Telangana. 19211a04d7@bvrit.ac.in N. Pranay Sai Reddy, Department of ECE, B V Raju Institute of Technology, Narsapur, Medak Dt, Telangana. 19211a04d8@bvrit.ac.in N. Shiva, Department of ECE, B V Raju Institute of Technology, Narsapur, Medak Dt, Telangana. 19211a04e0@bvrit.ac.in Abstract— The productivity of agriculture determines the economy of a country. To boost agricultural productivity and improve product quality, plant diseases must be identified. We can see agricultural production is rapidly growing and various techniques are being used for it. One of the emerging technologies is fogponics. In this technology, plants are grown using fog which is directly exposed to roots instead of water. This methodology has a limitation of being infected with diseases due to a lack of nutrition. Plant leaf patterns can be observed visually to diagnose disease using traditional methods; however, they require the use of human resources. Image classification is a technique to detect the diseases of plants that are already trained to model. If the disease is detected, then acknowledgment has been sent to the user that the plant is affected by the disease. This dataset must be uploaded to carry out the software implementation, and specific operations must be carried out on it to obtain the findings, namely whether the disease is detected. Regarding real-time applications, any model (Raspberry Pi) we use in our project struggles to hold substantial amounts of data. Devoid of any datasets for model training, we, therefore, apply a few different techniques to identify anomalous leaf conditions. A color feature of the leaf area is used in image-processing algorithms to detect plant infections and diseases. Keywords— Image Classification, Global System for Mobile Communication Module, Disease Detection, Raspberry Pi, Edge Impulse I. INTRODUCTION It is estimated that 70% of India's population relies on agriculture in 2021-2022. So, the production of agricultural products mostly affects the economy of a country more than other products. By 2018, the agriculture business section utilized more than 50% of Indian manpower and contributed 18% to national output [1]. The yield or crop's quality determines how it is used. A crop's nutrition is essential for assessing its quality and yield. There is still a possibility that a plant can get ill despite the nutrients being provided. We can say that the leaves of plants are the main indicator of disease when there is a problem with them. It is necessary to initially collect the required datasets of leaves of the plants that we are testing. However, the databases for plants and other classes are enormous. Therefore, it is exceedingly unlikely that a low-level processor could store all the data. Therefore, rather than training datasets and categorizing them, we favor live detection utilizing the Pi Camera. The next step is image pre-processing, and Grayscale images are made by converting RGB images into grayscale images using a color conversion technique. The image processing using a convolutional neural network achieves 97% accuracy of ResNet and 96.5% accuracy of AlexNet which are two models in CNN [2]. However, to understand the steps involved in the process we need to carry out the software implementation. We must first submit the datasets for specific classes of plants, such as tomatoes, potatoes, etc. All operations and algorithms are run on these photographs using a website called PLANT VILLAGE to create the dataset. A crucial step in clustering gray images is image segmentation, which involves categorizing them into 'n' - the number of pixels. This method can be carried out in different methods depending upon the type of crops or type of disease. Several methods use the K- Means Clustering Approach to divide the area affected by the lesion and combine global color histograms and color coherence vectors to create a local binary pattern that extracts color and texture features. In addition, SVM classifiers have an accuracy of 93% when there is more data, making them one of the best classifiers for classification. As part of their segmentation processes, several other methods mention threshold-based segmentation. In these methods, RGB image pixels are pixel values changed to zero when they are less than or equal to the threshold; If not, pixel values are adjusted to 255 and k-means clustering is carried out. To conclude, using the results of k-means, plot each pixel in the image. Different types of classes can be identified using this method [4]. SVM classifiers are used in this paper to classify data. And finally, after detecting and classifying the diseases the acknowledgement should be provided to the user’s mobile using GSM Module. II. LITERATURE SURVEY Using image processing, Chaitali G. Dhaware et al. [5] detected leaf disease in plants. Infectious and healthy images were used in a dataset containing 120 images. Pre-processing involves converting RGB images to HUE saturation values and resizing them to 512*512. With cluster subtraction, they were able to obtain results using image segmentation based on two techniques: cluster-based and color-based. This proposed model is for detecting different plant leaf diseases, but they have trained the model only for limited dataset images. Bhimte et al. [6] discuss how image processing can detect disease spots on cotton leaves. According to this paper, cotton leaves can suffer from bacterial blight, Alternaria leaf spots, and magnesium Proceedings of the International Conference on Sustainable Computing and Smart Systems (ICSCSS 2023) DVD Part Number: CFP23DJ3-DVD; ISBN: 979-8-3503-3359-6 979-8-3503-3360-2/23/$31.00 ©2023 IEEE 901