Automated Land Crude Oil Spill Detection with Gabor Filters and Wavelet Transform

Abstract

Abstract

Early detection of crude oil spills leads to significant reduction in the environmental pollution, economic losses and health hazards that result from these spills. The visiblelightspectrumstill remainsaresearcharea for spill detectionbecauseitis aquick economicalmethodformonitoringoil spills. This paper presents an algorithm for automated land crude oil spill detection in the visible spectrum. The algorithm employs three basic steps for crude oil spill detection: Gaborfiltration, wavelet transformanalysis, and color homogeneity extraction. The algorithm was tested on sixty (60) ground truth crude oil spill images obtained from the Crude Oil Spill Imaging Database (COSID). The algorithm was able to detect the crude oil spill in 55 of the 60 images, resulting in an efficiency of 91.7%.It was observed that in certain cases, the algorithm incorrectly flagged regions of vegetation as crude oil. The algorithm, therefore, requires the integration of a vegetation segmentation and removal step to reduce the number of false positives detection.

Keywords: Color homogeneity, crude oil spill detection, haar filters, Gaborfilters, wavelet analysis