Agriculture is dependent on the quality and quantity of plant growth worldwide. Many factors, such as weather, pest attack or disease, can cause plant-specific infections. The process of voluntary diagnosis is a long and laborious process, and the farmer cannot determine the cause and effect of the disease. This study proposes the optimal autocorrelation and classification of crop species (OADD-CS) using hybrid machine learning techniques. The first contribution is to propose a non-linear deep neural network (NL-DNN) for processing before removing abnormal image angles from the input image. Next, the edge target detection (ETD) algorithm is used to calculate the top and bottom leaf edges and extract the feature. Finally, hybrid crow-optimization-based convolutional neural network (HCO-CNN) classifiers are used to diagnose other diseases. Here we consider a dataset of over a million images taken with a mobile phone in real field conditions. This data set includes 17 diseases for 5 crops and 5 equally distributed diseases, with multiple diseases present in the same image. Specific work can be performed on MATLAB tools, and performance is compared to traditional complex work.