Monthly monitoring of a pipeline employing high spatial resolution images

Abstract

This study aimed to develop a methodology for the monthly monitoring of a pipeline using change detection techniques applied to a land cover mapping with high spatial resolution images. A 30.7km length and 400m width section of the Rio de Janeiro-Belo Horizonte pipeline route was selected in the cities of Duque de Caxias and Nova Iguaçu (RJ). The change detection process was developed using six GeoEye images from January to June 2010. An initial land cover mapping was obtained through visual interpretation of the January 2010 image. The multiresolution segmentation algorithm gives the object delimitation from the images grouped into five paired periods (Jan/Feb, Feb/Mar, Mar/Apr, Apr/May e May/Jun). A multitemporal algebraic procedure of the red bands from the images of each paired periods gives the changed areas. After this, the changed areas identified were classified through land cover, using fuzzy function. In the 11,890,000 m2 area section of pipeline analyzed, a total area of 775,402 m2 was detected as changed in six month period. In the changed area the main changes occurred for Bare Soil, Grassland and Cover/Soil ground. The classification process obtained an overall accuracy of between 71 and 80%. From these results, could be concluded that the procedure is appropriated and can contribute to the monitoring of pipeline zones. However, some of the changes are related to seasonal features (changes in humidity and sun elevation). Thus, to implement this monitoring method it is necessary to consider the influence of image acquisition parameters and perform experiments along a greater length of the pipeline in a different geographic environment.