OPTIMAL PARAMETER DETERMINATION FOR MEAN-SHIFT SEGMENTATION- BASED SHORELINE EXTRACTION USING LIDAR DATA, AERIAL ORTHOPHOTOS, AND SATELLITE IMAGERY

for shoreline extraction has been developed that is based on mean-shift segmentation and the integration of LiDAR data, satellite imagery and aerial orthophotos. This method first classifies LiDAR points as belonging either to a water surface or to land. The classification criterion is the homogenous nature of the Near-Infrared (N-IR) reflection of the water surface, the elevation, and color distribution. Subsequently a shoreline can be extracted by tracing the boundary between these two categories, water and land. However, each mean-shift process requires bandwidth to be used as the parameter to classify the dataset. Also, all data sources having different units of measurement need to be normalized. In this research, we focus primarily on the training phase of this method. Three parameters are necessary for the mean-shift bandwidth (one for each of the three classification stages). In addition, seven normalization scales are used for the data sources including 3-D coordinates (X, Y, Z), RGB values (R, G, B) and N-IR values in this classification phase. A small region of ground truth is needed to provide a reference for the classification performance in the training phase. We obtained this ground truth by manually classifying LiDAR points as either water or land and then manually tracing a shoreline from an orthophoto. When processing, the scale of elevation from the LiDAR points is set to one. Then a mean-shift algorithm with only the elevation feature is used to run the classification with the bandwidth from 0.1 m to 1 m in 0.1-meter steps. The classification result is then evaluated by looking at the rates of true positives, false positives, and false negatives and by the accuracy of the shoreline within the region of ground truth. The best bandwidth is selected based on statistical tests looking at the previously described evaluation factors. All other parameters and scales are resolved one-by-one in a similar manner. The parameters and scales resolved by this systematic training procedure can assure the accuracy and stability of the extracted shoreline.

[1]  Rongxing Li,et al.  Potential of high-resolution satellite imagery for national mapping products , 1998 .

[2]  Jon Sellars,et al.  A Sensor Fusion Approach to Coastal Mapping , 2005 .

[3]  Yizong Cheng,et al.  Mean Shift, Mode Seeking, and Clustering , 1995, IEEE Trans. Pattern Anal. Mach. Intell..

[4]  Kaichang Di,et al.  Digital Tide-Coordinated Shoreline , 2002 .

[5]  D. Whitman,et al.  Mapping Shoreline Position Using Airborne Laser Altimetry , 2004 .

[6]  Yun-Jae Choung,et al.  Extraction of blufflines from 2.5 dimensional Delaunay triangle mesh using LiDAR data , 2009 .

[7]  Shoreline Mapping from Airborne LIDAR in Shilshole Bay , Washington , 2003 .

[8]  Larry D. Hostetler,et al.  The estimation of the gradient of a density function, with applications in pattern recognition , 1975, IEEE Trans. Inf. Theory.

[9]  R. Holman,et al.  Estimation of Shoreline Position and Change using Airborne Topographic Lidar Data , 2002 .

[10]  Kaichang Di,et al.  AUTOMATIC SHORELINE EXTRACTION FROM HIGH-RESOLUTION IKONOS SATELLITE IMAGERY , 2003 .

[11]  J. Shan,et al.  Building boundary tracing and regularization from airborne lidar point clouds , 2007 .

[12]  Tarig Ali,et al.  New methods for positional quality assessment and change analysis of shoreline features , 2003 .

[13]  Rongxing Li,et al.  Estimation of blufflines using topographic lidar data and orthoimages , 2009 .

[14]  Kaichang Di,et al.  3-D Shoreline Extraction from IKONOS Satellite Imagery , 2003 .

[15]  Dorin Comaniciu,et al.  Mean Shift: A Robust Approach Toward Feature Space Analysis , 2002, IEEE Trans. Pattern Anal. Mach. Intell..

[16]  Aidong Zhang,et al.  Digitalization of coastal management and decision making supported by multi-dimensional geospatial information and analysis , 2006, DG.O.