Maize Plant diseases are the major concerns in the agricultural domain. But significant loss of yield occurs due to awful techniques. Hence automatic and accurate identification of the quantifying disease is very important. Most of the diseases symptoms are reflected in maize leaves, but diagnosis by experts in laboratories are costly and time-consuming. In this paper, Optimization of Convolution Neural Network is carried out by using Improved Gaussian Conducted Particle Swarm Optimization [IGCPSO] for classifying the maize diseases based on optimization. Initially Contrast Limited Adaptive Histogram Equalization [CLAHE] is framed for partitioning of designated image in to particular non-overlapping segments of similar sizes using deep learning architecture. Preprocessed image is segmented into several segments by employing graph cut segmentation process. Segmented disease regions are useful to extract the discriminative features for remaining process. Color Level Co-occurrence Matrix (CLCM) is a new texture based technique employed to extract informative features related to the diseased region. Improved Gaussian Conducted Particle Swarm optimization is emphasized to generate optimal features by identifying and clustering of important points having similar attributes for classification. In proposed model, experimental analysis using plant village dataset improves performance in terms of dice coefficient, sensitivity, and specificity towards disease identification.

Convolution Neural Network, Color Level Co-occurrence Matrix, Graph Cut Segmentation, Particle Swarm Optimization, Contrast Limited Adaptive Histogram Equalization