Empowering Agriculture: Classification of Diseases Affecting the Leaves and Spikes of Wheat Crop using Hybrid Deep Learning Methodology

Worldwide, wheat ranks third in both harvesting and consumption of grains. Nevertheless, illnesses ruin a significant portion of wheat crops. Wheat crops are vulnerable to more than twenty different diseases. Consequently, it becomes exceedingly difficult to manually diagnose these disorders. Increases in both production and quality can be achieved via the use of automated disease categorization in wheat. Moreover, it has the potential to be a valuable tool for evaluating crop quality and setting prices. Disease diagnosis and categorization can benefit from image analysis based on deep learning. Among wheat plants, the spike and leaves take the biggest hit. These features are diagnostic for the vast majority of illnesses. This is due to a combination of reasons, including the fact that farm laborers are often illiterate and the wide variety of agricultural goods available. There have been a number of different models put out there as possible answers to the problem of wheat harvest disease detection. In order to detect and classify illnesses that impact wheat harvests, this study presented a new approach called Hybrid Learning for Wheat Crop Disease Detection (HLWCDD). This methodology combines Convolutional Neural Network (CNN) and Random Forest (RF) algorithms. Deep Classification Learning Model (DCLM), an existing deep learning model, is essentially a hybrid of Neural Networks and Support Vector Machines (SVMs). The suggested model is compared to this. In this study, we assess both models and demonstrate how effective the suggested method is. The programme makes use of trained models to detect important features in the images. Using the main criteria mentioned earlier, the proposed approach can distinguish between wheat harvests that have been impacted by disease and those that have not. After collecting 3,200 photos for this investigation, the dependability of the findings was determined to be 97.29%. Out of the total number of photos, eleven classes showed sick crops and one showed healthy crops. The photos that make up the collection were rotated at different angles so that the proposed model could detect and categorize illnesses from multiple viewpoints.