Remote sensing image extraction of tidal channels based on Otsu and mathematical morphology

Abstract:

Tidal channel is one of the major tidal flat landforms. The Jiuduan Shoal in Yangtze River Estuary was taken as the object in this paper. Three different types of tidal channels in panchromatic band remote sensing data received from Landsat 8 satellite on February 15, 2015 were selected. Firstly, the top-hat transformation is to eliminate the problem of the backgrounds of uneven brightness accurately, if the method of maximum between-class variance(Otsu) is used directly; Then, Otsu is used to find an optimal threshold to make the maximum variance between tidal channel and background, obtaining a binary image; Thirdly, the broken tidal channel is connected by morphological dilation, and non-target is removed by morphological removal. Finally the skeleton of tidal channel is extracted and cropped, and the complete information of tidal channel is gained. Visual analysis and quantitative analysis are used to evaluate the accuracy of the information of tidal channel. The result shows that the method combining Otsu with mathematical morphology can be used to extract tidal channel information efficiently. The average accuracy can reach 93.0%, missing error and redundance error are 7.0% and 0.5% respectively.