Deep Learning Based Automated Wheat Disease Diagnosis System

Abstract

It\'scriticaltoidentifyandcategorize agricultural diseases. To preserve the quality and yield of crops. Crop disease identification using traditional methods takes a lot of time and labor. As a result, methods based on computers have beencreatedtoautomaticallydiagnosetheillness. In addition, behind rice and maize, wheat is the third most harvestedandconsumedgrain.Cropdiseasedetectionisoneof the most popular study subjects these days. Recently, several wheat illnesses have been recognized and categorized using deep learning algorithms. This article presents a Residual Network (ResNet152), regarded as one type of Convolutional Neural Network(CNN), anddeeplearning-basedapproach for identifying and categorizing wheat illnesses. Compared toother approaches now in use, the suggested method provides a greater level of accuracy in the identification and classification of various wheat illnesses. Furthermore, the results show that the suggested strategy offers early diagnosis and treatment of wheat illnesses, improving crop quality and output.

Introduction

Wheat, a staple crop feeding billions, faces threats from diseases that reduce yield and worsen global food insecurity. Early detection and management of wheat diseases are crucial to protect crop productivity. Advances in image processing, AI, machine learning, and deep learning offer promising tools for timely and accurate disease diagnosis, helping farmers prevent losses and manage crop health effectively over large areas.

Literature Review:
Researchers have developed advanced deep learning models, such as RFE-CNN, which combine convolutional neural networks (CNN) with techniques like residual channel attention and elliptic metric learning to improve wheat leaf disease identification. These models outperform traditional architectures like VGG-19, achieving high accuracy (up to 99.95%) on public datasets.

Existing Research:
The disease detection process involves collecting high-quality images of healthy and diseased wheat leaves, extracting important features to reduce redundancy, setting model parameters, and using deep CNN architectures like VGG19 for classification.

Proposed Work:
The approach includes gathering diverse data on wheat diseases, cleaning and augmenting images (rotating, flipping, scaling), normalizing data, splitting into training/validation/testing sets, and using a deep ResNet-152 model for classification. Residual blocks and skip connections in ResNet help train very deep networks effectively by avoiding vanishing gradients. The model is then evaluated on unseen data to verify accuracy and reliability.

Methodology:
High-quality datasets are preprocessed and augmented for robustness. CNNs, with layers such as convolutional, pooling, and fully connected layers, extract spatial features automatically. Transfer learning with pre-trained models (VGG, ResNet) speeds up training and improves accuracy. Performance metrics like accuracy, precision, recall, and F1 score assess the system, which is deployed via an interface for real-time disease diagnosis and continuously updated to maintain accuracy.

Results:
Using a publicly available dataset (LWDCD2020), which includes diseases like leaf rust, tan spot, and Fusarium head blight, the deep learning model successfully classifies wheat leaf conditions with strong predictive performance, enabling accurate disease detection in practical field conditions.

Conclusion

Deep learning models have developed into an efficient tool for precision agriculture, helping farmers to increase crop yields andmake well-informed decisions. This is due to ongoing advancementsin data collecting and the creation of diverse disease detection techniques. In general, the use of deep learning models in agriculture increasesfoodsecurity,lowerswaste,andenhancessustainabilityover time. Therefore, a deep learning model for identifying andcategorizingwheatillnessesis createdinthisarticle.Promisingresults areshown inthe detection and classification of several wheat illnesses by the suggested method utilizing the ResNet152 model. This can assist farmers in taking prompt action to prevent crop loss and guarantee a healthy harvest. Fast processing speed and great accuracy in identifying wheat crop diseases are just two of its many benefits. Whilethetrainingaccuracyis97.81%,thesuggested modelhasahigh testing accuracy of 93.27%. To evaluate the model\'s performance on bigger datasets and in various environmental settings, more investigation is necessary.

References

[1] Pawlak, Karolina. “Food security situation of selected highly developed countries againstdeveloping countries,” Journal of Agribusiness and RuralDevelopment,Vol.40,No.2,pp.385-398,2016. [2] Trivelli, Leonello, Andrea Apicella, Filippo Chiarello, Roberto Rana, Gualtiero Fantoni, and Angela Tarabella. “From precision agriculture to Industry 4.0: Unveiling technological connections in the agrifood sector,” British Food Journal, Vol. 121, No. 8, pp. 1730-1743, 2019. [3] Pranjali. B. Padol and A. A. Yadav, “SVM classifier based grape leaf disease detection,” IEEE Conference on Advances in Signal Processing, Pune, India,pp.175-179, 2016. Doi: 10.1109/CASP.2016.7746160. [4] Uzhinskiy, Alexander, Gennady Ososkov, Pavel Goncharov, and Andrey Nechaevskiy.“Multifunctional platform and mobile application for plant disease detection,” Center for European Union Research Workshop Proceeding, Vol. 2507, pp. 110- 114, 2019 [5] Lu, Jiang, Jie Hu, Guannan Zhao, Fenghua Mei, andChangshuiZhang.“Anin-fieldautomaticwheatdisease diagnosis system,” Computers andElectronics in Agriculture, Vol. 142, pp. 369-379, 2017. [6] Zhu,Peng,RoseAbramoff,DavidMakowski,and Philippe Ciais. “Uncovering the past and future climate drivers of wheat yield shocks in Europe with machine learning,” Earth\'s Future, Vol. 9, No. 5, pp. 1-20, 2021, doi: 10.1.29/2020EF001815. [7] Jahan, Nusrat, Paulo Flores, Zhaohui Liu, Andrew Friskop, Jithin Jose Mathew, and ZhaoZhang “Detecting and distinguishing wheat diseases using image processing and machine learning algorithms,” Annual international virtual meeting, American Society of Agricultural and Biological Engineers, pp. 1, 2020, doi:10.13031/202000372. [8] Azadbakht, Mohsen, Davoud Ashourloo, Hossein Aghighi, Soheil Radiom, and Abbas Alimohammadi. “Wheat leaf rust detection at canopy scale under different LAI levels using machine learning techniques,” Computers and Electronics in Agriculture, Vol. 156, pp. 119-128, 2019. [9] Khan, Habib, Ijaz Ul Haq, Muhammad Munsif, Shafi Ullah Khan, and Mi Young Lee. “Automated Wheat Diseases Classification Framework Using Advanced Machine Learning Technique,” Agriculture, Vol. 12, No. 8, pp. 1226, 2022. [10] Azimi, Nasrin, Omid Sofalian, Mahdi Davari, Ali Asghari, and Naser Zare. “Statistical and Machine Learning-Based FHB Detection in Durum Wheat,” Plant Breeding and Biotechnology, Vol. 8, No. 3, pp. 265-280, 2020 [11] Haider, Waleej, Aqeel-Ur Rehman, Nouman M. Durrani, and Sadiq Ur Rehman. “A Generic Approach for Wheat Disease Classification and Verification Using Expert Opinion for Knowledge-BasedDecisions,”IEEEAccess,vol.9,pp.31104-31129, 2021, doi: 10.1109/ACCESS.2021.3058582. [12] Zhang, Xin, Liangxiu Han, Yingying Dong, Yue Shi, Wenjiang Huang, Lianghao Han, Pablo González-Moreno, Huiqin Ma, Huichun Ye, andTam Sobeih. “ADeep Learning-Based Approach for Automated Yellow Rust Disease Detection from HighResolution Hyperspectral UAV Images,” MDPI Journal,pp.1-13,June2019.DOI:10.3390/rs11131554 [13] Hasan, Md Mehedi, Joshua P. Chopin, Hamid Laga, and Stanley J. Miklavcic. “Detection andanalysisofwheatspikesusingconvolutionalneuralnetworks,”PlantMethods,Vol.14,pp.1-13, 2018. [14] Xu,Laixiang,BingxuCao,FengjieZhao, Shiyuan Ning, Peng Xu, Wenbo Zhang, andXiangguan Hou. “Wheat leaf disease identification basedondeeplearningalgorithms,”Physiologicaland Molecular Plant Pathology, Vol. 123, pp. 101940, 2023. doi.org/10.1016/j.pmpp.2022.101940

Copyright

Copyright © 2025 Shri Rama Chandra Murthy, Samboji Shyam Kishor, Sagi Sashi Sekhar Varma , Dr. Ch. Srinivasulu. 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.