10:45 AM - 11:00 AM
[STT36-07] Examination of extraction method of topographical change due to eruption of Nishinoshima using time series SAR
Keywords:time series SAR, Nishinoshima
The SAR image is subjected to two processes, synthetic aperture technology and pulse compression technology, to form a grid-like mesh on the ground surface, and the intensity of reflected waves is measured for each pixel. The reflected wave also differs greatly depending on the object. Therefore, in order to classify the land area and the sea area from the images observed by SAR, we consider the definition of the land area and the sea area. The land area is a place where the altitude is 0 m or more, and the backscattered signal in the land area is generally larger than that in the sea area. On the other hand, the sea area is an altitude of 0 m, and the reflected signal is generally small. Since the signal in the land area is larger than that in the sea area, it looks bright in the amplitude image.
In this study, we study Nishinoshima, an uninhabited volcanic island in the Ogasawara Islands of Japan, using time-series SAR data, focusing on temporal changes and area changes in topography associated with eruptions.In addition, the accuracy of the results obtained by these methods will be evaluated. Among the topographical changes, three methods were implemented for the area. The data used is time-series data of 23 scenes observed by JAXA's ALOS-2 / PALSAR-2 for 6 years from September 29, 2014.
The first is the threshold method. Determine one optimal threshold as a whole. The area is calculated by counting the number of pixels above the threshold value as the number of pixels on land. On the other hand, as the island expands, the brightness changes depending on the location due to the effects of ash fall. Therefore, the image is divided into four parts, threshold values are set in each area, and the land area is calculated. Therefore, the image is divided into four parts, threshold values are set in each area, and the land area is calculated. All of them are added together to calculate the total land area.
The second is the automation method. This method uses image analysis software called ImageJ to automatically determine the threshold. As a result, an image in which the land is white and the sea is black is created, and the area is calculated.
The third is the edge detection method. By multiplying the original image by the convolution matrix in the vertical and horizontal directions and adding the obtained pixel values, a differential weighted image with emphasized boundaries is created. By multiplying the original image by the convolution matrix in the vertical and horizontal directions and adding the obtained pixel values, a differential emphasis image with emphasized boundaries is created. This is an image with an increased backscattering coefficient on land. Therefore, even in the image where the backscattering coefficient of a part of the land is lowered due to the accumulation of volcanic ash, the backscattering coefficient of the sea is larger than the backscattering coefficient of the land as a whole, and the backscattering coefficient of the coastal area changes slightly can be extracted.
As a result, with respect to the local data of the Japan Coast Guard and the Japan Meteorological Agency, the relative error of each method is -2.79% for the threshold method, -12.63% for the threshold method when the island is divided into four, and -10.94% for the automation method,the edge detection method resulted in -5.56%. At this point, the threshold method is considered to be the most effective for extracting the area of the island, and the edge detection method may also be an effective means depending on the treatment of sea noise.
In this study, we study Nishinoshima, an uninhabited volcanic island in the Ogasawara Islands of Japan, using time-series SAR data, focusing on temporal changes and area changes in topography associated with eruptions.In addition, the accuracy of the results obtained by these methods will be evaluated. Among the topographical changes, three methods were implemented for the area. The data used is time-series data of 23 scenes observed by JAXA's ALOS-2 / PALSAR-2 for 6 years from September 29, 2014.
The first is the threshold method. Determine one optimal threshold as a whole. The area is calculated by counting the number of pixels above the threshold value as the number of pixels on land. On the other hand, as the island expands, the brightness changes depending on the location due to the effects of ash fall. Therefore, the image is divided into four parts, threshold values are set in each area, and the land area is calculated. Therefore, the image is divided into four parts, threshold values are set in each area, and the land area is calculated. All of them are added together to calculate the total land area.
The second is the automation method. This method uses image analysis software called ImageJ to automatically determine the threshold. As a result, an image in which the land is white and the sea is black is created, and the area is calculated.
The third is the edge detection method. By multiplying the original image by the convolution matrix in the vertical and horizontal directions and adding the obtained pixel values, a differential weighted image with emphasized boundaries is created. By multiplying the original image by the convolution matrix in the vertical and horizontal directions and adding the obtained pixel values, a differential emphasis image with emphasized boundaries is created. This is an image with an increased backscattering coefficient on land. Therefore, even in the image where the backscattering coefficient of a part of the land is lowered due to the accumulation of volcanic ash, the backscattering coefficient of the sea is larger than the backscattering coefficient of the land as a whole, and the backscattering coefficient of the coastal area changes slightly can be extracted.
As a result, with respect to the local data of the Japan Coast Guard and the Japan Meteorological Agency, the relative error of each method is -2.79% for the threshold method, -12.63% for the threshold method when the island is divided into four, and -10.94% for the automation method,the edge detection method resulted in -5.56%. At this point, the threshold method is considered to be the most effective for extracting the area of the island, and the edge detection method may also be an effective means depending on the treatment of sea noise.