Soil category is one of most significant factors which aids to decide that which type of crop should be planted in order to get effective yield. Common methods used by the farmers are in adequate for fulfilling the growingdemands and thus they have to hamper the cultivating soil. Agriculturalists must aware of suitable soil types for a specific crop, for proper crop yield and which would affect the increasing demand for food. There exists several laboratory and field techniques for classifying soil, however these have limitations namely labour-intensiveand time consuming. There occurs a need of computer-based soil classification methods, which will aid farmers timely and accurately. This article develops an automated soil classification using red deer optimization with deep belief network (ASC-RDODBN) model. The presented ASC-RDODBN model mainly aims to recognize different kinds of soil, which helps in proper crop type mapping process. To attain this, the presented ASC-RDODBN model undergoes two stages of preprocessing such as bilateral filtering (BF) based noise elimination and contrast enhancement. In addition, data augmentation process is performed to increase the dataset size. For feature extraction, the ASC-RDODBN algorithm uses radiomics features. Finally, DBN model is applied for soil classification process in which the DBN hyperparameters are tuned properly using RDOmodel. Extensive simulations were carried out on soil type classification dataset from Kaggle repository to highlight the better performance of the ASC-RDODBN model. The experimental outcome pointed out the enhancement of the presented ASC-RDODBN model over recent models.