A Survey on Deep Learning Technique for Sugarcane Disease Detection

Abstract

Sugarcane is a vital crop, and its health directly impacts agricultural yields and the sugar industry. To address the challenges associated with disease detection in sugarcane, we propose an approach using deep learning techniques. Our study leverages convolutional neural networks (CNNs) and image analysis to accurately identify and classify various sugarcane diseases. By analyzing high-resolution images of sugarcane leaves and stems, our deep learning algorithm provides remarkable accuracy in disease detection, offering a promising solution for early diagnosis. This research contributes to sustainable agriculture and aids in preserving the economic viability of sugarcane cultivation.

Introduction

I. INTRODUCTION

Sugarcane is a vital crop that is essential to the global agriculture industry, serving as a primary source of sugar production and bioenergy. However, the cultivation of sugarcane is constantly challenged by the appearance of numerous diseases that can significantly reduce yield and quality. Timely and accurate disease detection is necessary for sustainable sugarcane farming and efficient disease control. Traditional methods of disease diagnosis in sugarcane rely on visual inspection by experts, which may require a lot of time, subjective and prone to errors. To address these challenges, the integration of deep learning algorithms has emerged as a promising solution for automating and increasing the accuracy of sugarcane disease detection.

In recent years, deep learning has revolutionized the field of computer vision, enabling machines to automatically learn and identify complex patterns and features within images. This technology has found applications in various domains, including agriculture, where it has the potential to enhance crop disease detection. Deep learning algorithms can be used to build reliable and effective systems for the automatic diagnosis of sugarcane diseases, thus reducing the reliance on manual inspection and improving the overall health and productivity of sugarcane crops. In this paper, we delve into the exciting field of sugarcane disease detection using deep learning. We explore the challenges associated with traditional disease diagnosis methods, highlight the advantages of deep learning, and present recent developments in the application of this technology for identifying sugarcane diseases. Furthermore, we discuss the potential impact of this innovative approach on the sugarcane industry, emphasizing the benefits of early disease detection, improved resource management, and enhanced crop sustainability. By leveraging the capabilities of deep learning, we aim to pave the way for a more efficient, accurate and sustainable future in sugarcane cultivation.

 II. PROBLEM STATEMENT

The cultivation of sugarcane plays a pivotal role in the global sugar industry, contributing significantly to the world's sugar supply. However, the occurrence of various diseases in sugarcane plants poses a significant threat to crop yield and quality. Detecting these diseases early is crucial for effective disease management, but traditional methods are often labor-intensive and time-consuming. This project aims to deal with the issue by developing a deep learning-based solution for the automated and accurate detection of sugarcane diseases. The challenge is to produce a robust and effective model that can categorize and identify these diseases in real-time, aiding farmers in timely intervention and minimizing crop losses.

III. EXISTING SYSTEM

The Sugarcane Disease Detection system makes use of machine learning methods to identify and classify diseases affecting sugarcane crops. It relies on image recognition algorithms that analyze images of sugarcane leaves to identify common diseases like rust, smut, and mosaic virus. The system employs a dataset of labeled images for training and validation, and it continuously learns and updates its disease detection models. This technology aids farmers to detect diseases in advance, allowing for timely intervention and improved crop yield. It contributes to sustainable agriculture by reducing the usage of pesticides and enhancing crop management practices.

IV.  RELATED WORKS

Sugarcane Disease Detection Using CNN Deep Learning: An Indian Perspective - In the paper [1], the design uses a simplified convolutional neural network with four different classes to detect sugarcane diseases with 98.69% accuracy. The trained model was successful in recognizing and classifying sugarcane images into diseased/diseased and non-diseased/healthy groups based on sugarcane leaf pattern and disease. Thanks to this research, farmers can identify and classify sugarcane diseases using computer vision and machine learning. The main objective of the research presented here is to provide inputs to CNN, which would provide essential assistance to farmers through a web-based application that can be accessed from any web-connected device (e.g., desktop, mobile, tab, etc.)  needs. A trained model on real-time images can be well used in detecting and identifying plant diseases. The proposed system is a web application. Thus, we can log into the system and easily enter the image into the classifier to get a result that makes the system more user-friendly. The usefulness and application of this technique depends on how well the model can be adapted to real-world use through a web-based application. Customization based on user feedback of the model can be considered as a future area of this research. An adaptive and continuously updated robust dataset based on user input is the key to the utility and application improvement of this technology. This can be considered as a limitation of the study and can be considered as a future area for other researchers along with updating the dataset in an iterative mode while the web-based application is running.

1. Classification of Sugarcane Leaf Diseases using Deep Learning Algorithms: The paper [2] shows how to categorize sugarcane leaf diseases using deep learning techniques. The models demonstrated an overall accuracy of 99.24% on average, according to the results. The model based on InceptionV3 achieved the greatest accuracy of 99.53 during training, 99.61% during validation, and 99.70% during testing. Thus, our work validates the utility of deep learning-based models for sugarcane leaf disease categorization. This study does have certain drawbacks, though. At first, this work was done using a small number of data sets. Therefore, more data can be used in future research to improve the models' performance. Secondly, it is possible to fine-tune other models that were utilized in this experiment to attain the best possible result. To collect valuable data in real-world scenarios, a testing activity intended for model integration and deployment with systems should be completed.
2. Research Paper on Sugarcane Disease Detection Model: The paper [3] adopts an approach based on image processing combined with deep learning (convolutional neural networks) to first identify and then classify leaves according to their diseases. Data acquisition is done considering versions with Lehkuva's RGB filter. In the pre-processing phase, the noise of the images was removed with filters. The extraction of image features is done so that the features of the disease symptoms from the review journal are obtained. Image classification is done using Convolutional Neural Networks (CNN). Based on our experiment, we achieved the required results with satisfactory accuracy and precision due to the effective nature of CNN in this image classification task.
3. Sugarcane Disease Detection using Deep Learning: The paper [4] applies deep learning algorithms to detect and classify whether a sugarcane leaf is diseased or healthy. The architecture used a basic convolutional neural network with 7 different classes to categorize the sugarcane leaf and achieved an accuracy of 95%. The trained model achieved its goal by effectively identifying and classifying sugarcane images into healthy and diseased categories depending on the leaf pattern. Therefore, this study provides ideas to help farmers identify and classify sugarcane diseases using computer vision and machine learning. For future work, different models can be applied to determine the performance of the model on the training set. Experiments with the suggested models can also be conducted using various optimizers and learning rates.
4. Deep Learning in Agriculture: A Survey: The paper [5] conducted a research on applied research activities based on deep learning in the sphere of agriculture. It identifies 40 relevant articles dealing with a specific issue the domain and problem they focus on, the specifications of the models they use, the data sources used pre-processing tasks and data techniques and in general performance of each paper and performance indicators. Deep learning was then compared to with respect to the performance with other existing techniques. The findings in the United States indicate that deep learning performs better and surpasses other widely used image processing methods. Future works, intend to apply a general deep learning concepts and best practices as described in this study in agricultural sectors where this modern technology has not yet been sufficiently used. The goal is that this study will encourage more researchers to try in more detail learning, its application to various agricultural problems involving classification or computer vision and image analysis or, more generally, data-driven forecasting analysis. The overall benefits of deep learning encourage its continued use smarter, more sustainable agriculture and safer food production.
5. Deep Neural Networks Based Recognition of Plant Diseases by Leaf Image Classification: The paper [6] investigated a new means to use a deep learning method for classification and identification of plant diseases based on leaf images. There are numerous methods in the automated process of identification and categorizing of plant diseases, but this area of research is lacking. Furthermore, there is lack of profitable solutions on the market other than plant species identification based on leaf pictures.

The developed model was capable to identify the presence of leaves and differentiate between healthy and unhealthy leaves that can be diagnosed visually. Various experiments were carried out to verify the conduct of the new model. A new plant disease dataset was created, containing certain number of images taken from available online sources and expanded to even more number of images with suitable variations. Improvements did not show significant changes in overall accuracy, but the addition process had a greater impact on achieving respectable results. Since the presented method cannot be used in the field of detection of plant diseases, no comparison was made with similar results using precise technology. An extended version of this research concerns image collection to enhance the dataset and improve model accuracy using various fine-tuning and upscaling techniques. This program helps agriculturists to quickly and efficiently recognize plant diseases and facilitate making decisions while using chemical pesticides. In addition, future enhancements will expand the usage of the trained model to recognize plant diseases over larger land. By expanding this analysis, the authors desire to make a helpful significance on sustainability by influencing the quality of crops for generations to come.

V. OBJECTIVES

1. Develop a deep learning algorithm that can accurately identify and classify various diseases affecting sugarcane crops.
2. Create a comprehensive dataset of sugarcane plant images, including healthy plants and plants with different disease symptoms, to instruct and evaluate the deep learning model.
3. Use pre-processing methods like image normalization, augmentation, and noise reduction to improve the consistency and quality of the image data.
4. Investigate various deep learning algorithms like Convolutional Neural Networks (CNNs), to discover the most effective model for disease detection in sugarcane plants.
5. Optimize hyperparameters, including learning rates, batch sizes, and network architecture, to improve the model's performance in terms of accuracy and efficiency.
6. Evaluate the model's performance through cross-validation and other relevant metrics to ensure its generalization and reliability in real-world scenarios.
7. Develop a user-friendly interface or application that can take input pictures of sugarcane plants and provide real-time disease detection results.
8. Explore the integration of remote sensing and drone technology for the automated capture of images from sugarcane fields for disease monitoring.
9. Investigate the potential for early disease detection using spectral analysis and multispectral imaging techniques.
10. Collaborate with agricultural experts and farmers to gather domain-specific knowledge, validate the model's results, and provide actionable insights for disease management and crop protection.
11. Create educational materials and resources to raise awareness among farmers about sugarcane diseases.
12. Document the research and development process in scientific publications and presentations to contribute to the broader field of agricultural technology and deep learning applications.
13. Continuously update and improve the deep learning model to adapt to new disease strains and variations, ensuring its long-term utility in the sugarcane farming industry.

VI. METHODOLOGY

A. Flow Diagram

1. Dataset Collection

• Collect images of sugarcane plants with various diseases and healthy ones. These images should be labeled to indicate the presence of disease or health.

2. Data Preprocessing:

• Cleaning, normalization and transformation of collected images.
• The material was divided into test and training sets.

3. Image Segmentation:

• Segment images of sugarcane plants to separate regions of interest (stem and leaves).
• This step helps to focus on specific parts of the plant where diseases can occur.

4. Feature Extraction:

• Extract relevant features from segmented images. These features may include color histograms, textures or deep learning-based features such as convolutional neural network (CNN) activations.

5. Deep Learning Model:

• Build and train a deep learning model, typically a CNN, to learn patterns and features from segmented images.
• The model should be designed to classify images into disease categories (e.g., different sugarcane diseases or healthy ones).

6. Classification of Disease

• Use a trained deep learning algorithm to classify segmented images into disease or health categories.
• The algorithm gives each class's probability scores.

7. Result Analysis:

• Analyze results including accuracy, precision, recall and F1 score to assess the level of performance of the disease detection model.
• Visualize the results to understand the distribution of diseases in the data set and the pattern and predictions.

B. Procedure

1. Problem Definition: Clearly articulate the problem to solve, how to use deep learning algorithms to detect diseases in sugarcane. Set goals and expected results.
2. Data Collection: Collect a diverse and comprehensive dataset of images of sugarcane cultivation. This dataset should include images of healthy sugarcane plants and plants suffering from various diseases. Make sure the dataset reflects the different stages and severities of these diseases.
3. Data Preprocessing: Clean and preprocess the data set. This step involves resizing images, removing irrelevant data and expanding the dataset to increase its size. Data augmentation approaches such as rotation and color correction can make the model more robust.
4. Data Labeling: Label the dataset with accurate disease tags. Each picture must indicate the corresponding type of disease and the degree of severity. This is an important step in guided learning.
5. Data Partitioning: Divide the data set into three subsets: training, validation and testing. A common split ratio is 70-15-15 or 80-10-10.
6. Model Selection: Choose the best deep learning architecture for the job. Applications of Convolutional Neural Networks (CNNs) for image categorization are common. Consider using pre-trained models that can be adapted to your situation, such as VGG, ResNet or Inception.
7. Model Development: Develop a trained model by designing a neural network architecture. Make sure it can handle the dimensions of the provided image and print the degree and severity of the disease. Implement your chosen architecture utilizing a deep learning structure such as TensorFlow or PyTorch.
8. Model Training: Train the model on the training dataset. Use appropriate loss functions, optimization techniques (e.g., Adam, SGD) and hyperparameters. Track the training process and performance in the validation set. Use techniques such as early stopping to prevent overfitting.
9. Model Evaluation: Evaluate the trained model and its performance on the test image database using relevant metrics such as accuracy, precision, recall, F1 scores and confusion matrices. Assess its capability to accurately detect and determine the severity of the disease.
10. Hyperparameter Tuning: Fine tune hyperparameters to optimize model and performance. This may include adjusting learning rates, batch sizes, or other architectural elements.
11. Interpret the Model: Analyze the model and predictions and understand what features it uses to make decisions. Imaging techniques such as Grad-CAM or activation maps can help interpret the model.
12. Deployment: Deploy the trained model in actual environment, like mobile application or website, to detect diseases on the spot.
13. Monitoring and Maintenance: Continuously monitor the model and its performance in the field and instruct it with new data as needed. Update the system to match changing disease or environmental conditions.
14. Documentation: Document all steps, assemblies and results clearly and systematically so that others can understand and replicate your work.
15. Publishing and Sharing: Share your results and methodology in the manner of research papers or presentations, contributing to the agricultural knowledge base.

VII. ACKNOWLEDGMENT

We would like to express our deep gratitude to Mr. KUMAR K for his valuable and constructive suggestions during the planning and development of this project. His willingness to give his time so generously has been very much appreciated. We would also like to thank all the professors of KSIT for their continuous support and encouragement.

Conclusion

In conclusion, the application of deep learning in sugar cane disease detection represents a groundbreaking advancement in agriculture and crop management. The use of neural networks and image analysis has demonstrated remarkable accuracy in identifying and classifying diseases affecting sugar cane crops. This technology promises to revolutionize the way farmers monitor and protect their fields, enabling early detection, targeted intervention, and the potential for increased yields and sustainability. The benefits extend beyond individual farmers to the broader agricultural community and food security. However, continued research and refinement are essential to harness the full potential of this technology, ensuring its successful integration into agricultural practices. Ultimately, the convergence of deep learning and agriculture holds great promise for more efficient and sustainable crop production in a rapidly evolving world.

References

[1] Sammed Abhinandan Upadhye, Maneetkumar Rangnzath Dhanvijay, Sudhir Madhav Patil, \"Sugarcane Disease Detection Using CNN-Deep Learning Method: An Indian Perspective,\" Universal Journal of Agricultural Research, Vol. 11, No. 1, pp. 80 - 97, 2023. [2] Sammy V. Militante1 , Bobby D. Gerardo? , Ruji P. Medina, Sugarcane Disease Recognition using Deep Learning , 2019 IEEE Eurasia Conference on IOT, Communication and Engineering, October 2019 [3] S. Militante, Fruit Grading of Garcinia Binucao (Batuan) using Image Processing, International Journal of Recent Technology and Engineering (IJRTE), vol. 8 issue 2, pp. 1829- 1832, 2019. [4] A. Kamilaris and F. X. Prenafeta-Boldú, Deep learning in agriculture: A survey, Computer Electronics Agriculture, vol. 147, no. July 2017, pp. 70–90, 2018 [5] H. Park, J. S. Eun and S. H. Kim, Image-based disease diagnosing and predicting of the crops through the deep learning mechanism, In Information and Communication Technology Convergence (ICTC), IEEE 2017 International Conference on, pp. 129-131, 2017. [6] J. Amara, B. Bouaziz, and A. Algergawy, A Deep Learning-based Approach for Banana Leaf Diseases Classification, BTW, pp. 79–88, 2017. [7] S.H. Lee, C.S. Chan, P. Wilkin, P. Remagnino, Deep-plant: Plant identification with convolutional neural networks, 2015 IEEE Intl Conf. on Image Processing, pp. 452–456, 2015. [8] Kaur, Iqbaldeep, Gifty Aggarwal, and Amit Verma, “Detection and Classification of Disease Affected Region of Plant Leaves using Image Processing Technique,” Indian Journal of Science and Technology, vol. 9, no. 48, 2016.

Copyright

Copyright © 2023 Mr. Kumar K, N Keerthana, Sahana R, Shilpa M, Vandana N. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.