Evaluating methods to classify sugarcane planting using convolutional neural network and random forest algorithms
International Journal of Development Research
Evaluating methods to classify sugarcane planting using convolutional neural network and random forest algorithms
Received 14th September, 2020; Received in revised form 06th October, 2020; Accepted 29th November, 2020; Published online 30th December, 2020
Copyright © 2020, Flávio Henrique dos Santos Silva et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
This work aimed to develop an algorithm based on Convolutional Neural Networks and to compare it to a GIS plugin for automatic classification of sugarcane planting areas in medium spatial resolution images from Remote Sensing (RS) at municipality of Coruripe/AL, Brazil. We used Qgis software to process Land Remote Sensing Satellite (Landsat) images and to calculate sugarcane class areas in 2018, followed by Python, OpenCV, Keras and Tensorflow for developing algorithm to images training and classification, and finally, used Random Forest Classifier (RF) algorithm. Neural network algorithm classified samples obtained from 2018 image and mapped 172,65 km² of Sugarcane class and 71,71 km² of Not Sugarcane class. Já para classificação utilizando as amostras de 1986 + 2018 delimitou 155,95 km² de classe SUGARCANE e 88,41 km² como NOT SUGARCANE. Both Neural Network and Random Forest algorithms presented similar results, mainly in relation to the value of the area classified as Sugarcane and Not Sugarcane. Random Forest method performance for classification proven can validate algorithms developed in convolutional neural network for Remote Sensing data classification.
