Optimal deep convolutional neural network based crop classification model on multispectral remote sensing images.

• Propose a new ODCNN-CCM model for crop classification on multispectral remote sensing images. • Employ WF based image preprocessing and RetinaNet model as a feature extractor. • Apply DSO with DSAE model is applied for the crop classification process. Multispectral remote sensing images (MRSI)are widely employed to assess modifications in water bodies, land use and land cover changes, forest degradation, landscape change, and so on. Conventionally, MRSI is applied to map crops globally. MRSI based crop classification has gained considerable interest in the areas of agricultural productivity, agricultural policies, assuring food security, and recognizing sustainable agricultural development. Recently developed deep learning models can also be employed for crop classification using multispectral remote sensing images. In this aspect, this paper presents an optimal deep convolutional neural network based crop classification model (ODCNN-CCM) using multispectral remote sensing images. The presented ODCNN-CCM technique initially employs adaptive wiener filtering based image pre-processing technique. Moreover, RetinaNet model is applied to perform feature extraction process. Finally, dolphin swarm optimization (DSO) with deep stacked denoising autoencoder (DSDAE) model is applied for crop type classification. The performance validation of the proposed technique is validated using the Indian Pines Benchmark (INB), University of Pavia Benchmark (UPB), and Salinas Scene Benchmark (SSB). The proposed model achieves maximum accuracy of 97.51%, 98.33%, and 97.75% on the INB, UPB, and SSD datasets respectively. [ABSTRACT FROM AUTHOR]