A comparison of pixel and object-based change detection techniques using Landsat-8 multispectral imagery from Barrancabermeja city

Remote sensing techniques collect relevant information from Earth surface. This information is useful for deforestation analysis, natural disasters monitoring, urban expansion analysis, and study of climate changes. Multispectral and hyperspectral systems had powered this applications allowing temporal analysis using classical pixel-based techniques and most promising techniques like object-based classifiers. Aiming to study land cover changes in Colombian territory, a remote sensing change detection study is presented in this work. A comparison between pixel-based and object-based techniques was performed over multispectral imaging from Landsat-8 collected over Barrancabermeja city. Obtained results show that object-based classification technique outperform the pixel based one, besides it shows that Barrancabermeja region is constantly changing, specially by the crops in the rural area but slightly affected by deforestation.