Pigeon Pea Leaf Illness Identification and Categorization Using Bayesian Optimizer Deep Hybrid Learning Approach

Vanita Bhimappa Doddamani, Geeta Babusingh, G G Rajput

Abstract: The pigeon pea, or Cajanus cajan, is a significant legume crop valued for both its commercial value and high nutritional content. However, there are certain diseases that could impact its output and quality. Accurate and quick identification is essential for managing and controlling these disorders. Without expert help, the time-consuming and tedious method of manually identifying pigeon pea leaf disorder could produce unreliable findings. For them to guarantee the purity and quantity of pigeon pea production by offering prompt treatments to reduce disease transmission, farmers must have an automated, early, and accurate leaf disease recognition system. A Bayesian optimized deep hybrid learning system evolved for identifying and classifying damages in pigeon pea leaves. Standardizing the data through substantial preprocessing is necessary to build a custom dataset pigeon pea leaf. To improve the model's speed and accuracy, leaves are separated from its complex background using a modified U-Net segmentation technique. Features are extracted from CNN to get relevant features and then various machine learning classifiers are used for classification. On an unseen dataset, this model achieved accuracy values of 99.5%. The results reveal that the suggested model provides an effective solution for detection and classification in agricultural application in identifying illnesses in pigeon pea leaves.

Keywords: Pigeon pea leaves Disease detection, Image classification, segmentation, Agricultural applications