Super Resolved Maize Plant Leaves Disease Detection Using Optimal Generative Adversarial Network

Agriculture plays a vital role in the economy and crop disease causes huge financial losses every year. The losses can be reduced by detecting the disease accurately. The variation in light intensity and complex background of the agricultural field in detecting the maize leaves disease are the biggest challenges. An optimization algorithm, named Cat Swarm Political Optimizer Algorithm (CSPOA) has been developed in this research to detect the disease of a maize plant leaf. Our proposed algorithm is an integration of the Cat Swarm Optimization (CSO) and Political Optimizer (PO) algorithm. Anisotropic filtering performs pre-processing for removing noise and the Region of Interest (ROI) extraction for enhancing the image quality. The super resolution image is obtained from the Low Resolution (LR) images using kernel regression model. After obtaining the super resolution image, the salient map extraction has been carried out for representing the saliency. Finally, the maize plant leaves disease classification process is done using General Adversarial Network (GAN) for identifying the maize leaves disease. The training of GAN develops the CSPOA. On comparing with the existing maize plant leaves disease detection approaches, the developed CSPOA-based GAN performed with a maximum accuracy 0.9056, maximum sensitivity 0.9599, and the maximum specificity 0.9592, respectively.