remote sensing of crops

Deep learning Convolutional Neural Network (CNN) for Cotton, Mulberry and Sugarcane Classification using Hyperspectral Remote Sensing Data

Kavita Bhosle, Bhakti Ahirwadkar


Crop Classification using remote sensing data is important for calculating crop sown area and predicting the crop production. Accuracy in data will help to regulate marketing of the produce. Present study aims to examine the use of deep learning convolutional neural network (CNN) to overcome the difficulties arising in crop identification with satellite images. In the present work, EO-1 Hyperion hyperspectral images have been used for identifying cotton, sugarcane and mulberry crop. Structured data has been extracted from hyperspectral data for performing experiments. Deep learning convolutional neural network (CNN) is compared with deep feed forward neural network (FFNN). It is observed that, deep learning CNN provided 99.33 % accuracy, while deep FFNN gave 96.6 % accuracy. Empirical results demonstrate that CNN works well in practice and compares appreciatively to deep FFNN methods. Moreover, deep learning CNN has demonstrated efficiently for smaller size dataset.


Remote sensing data, Convolutional neural network, Principal component analysis, Hyperspectral data, Deep learning, Deep Feed Forward Neural Network

Full Text: