Skip to Local Navigation
Skip to Content
California State University, Long Beach
CSULB Geospatial Research and Mapping (GRAM) Field Program
Print this pageAdd this page to your favoritesSelect a font sizeSelect a small fontSelect a medium fontSelect a large font

Lawrence Fujiwara – Analyzing historic Hoouluia heiau through UAV imagery and OBIA methodology

Analyzing historic Hoouluia heiau through UAV imagery and OBIA methodology

Lawrence Fujiwara
California State University, Long Beach


The purpose of this research was to find a way to detect and define archaeological structures by analyzing and observing patterns in the scale of individual rocks using remote sensing techniques. The focus of this approach is a Heiau located on the southeast coast of Kauai (Figure 1). I investigated the ability to use remote sensing techniques to quantify differences in the structure of rocks used to build this archaeological feature. 

Figure 1. Location of the study area


1. A kite with Cannon G12 camera was flown over the heiau to take aerial photography.
2.The images collected were photo mosaicked by using Agisoft photoscan.
3.The mosaiced photos were georeferenced with ground control points collected using a Trimble GeoXH GPS.
4.Using eCognition, individual rocks were converted to polygon features with associated attributes including brightness, compactness, area (sq. meter) and others. (Figure 2).
Figure 2. eCognition segmentation of the heiau.. Scale parameter 10, shape 0.3, compactness 1.
5.These polygon features were then exported to ArcMap.
6.Using the LIDAR data purchased by CSULB, I made a slope, hillshade and contour lines of the heiau to assess topographic structure of the feature. (Figure 3 and 4). Using this information, the lower slope and the flat platform of the heiau were digitized. (Figure 5).
Figure 3. Map of hillshade and contour lines of the heiau created from the LIDAR data.Figure 4. Slope of  heiau created from the LIDAR data.

7.Used select by location to select polygons produced from eCogntion that intersected within the digitized polygons for both flat and sloped area in the heiau. (Figure 6).
8.Attribute table of both flat and slopped area were exported into an excel file.
Figure 5. Digitized layers of both flat and sloped area . Both are separate layers and 15% transparent.*While digitizing I made the slope and aerial image of the heiau transparent. I also made sure not to include vegetation and rocks that are not part of the heiau.

Figure 6. Screen shot of the two layers after select by location with the polygons produced from eCognition.


9.IBM SPSS was used to perform an independent sample t-test to assess to see if there is a statistically significant difference between rocks on the flat part .
10.The attributes used for the t-test was compactness, brightness, elliptic fit, main direction, density, radius and surface area in sq.meter.


With the merged layers of both flat and sloped area, there was a definite pattern in the heiau. The rocks on the flat area of heiau had smaller surface area (sq.meters) (Figure 7).

Figure 7. Map of surface area of both flat and sloped . The flat and  lowerslope polygons were merged. The values are in square meters.

The results of the sample t-test which compares the mean score of flat and sloped area showed that compactness, brightness, elliptic fit, density, radius and surface area were all highly significant at the 0.05 significant level. However it was not significant in the direction. (Table 1 and 2).

This confirms rocks in the flat and sloped area are quite different in many ways.

Table 1. Statistic outputs of IBM SPSS

Table 2. Statistic outputs of IBM SPSS


The result of this analysis suggests that the larger rocks found on the slope maybe part of the internal composition of the deteriorating heiau due to natural causes. More research needs to be done on other heiaus and/or other archaeological structures. With more data on the patterns of rocks, size of rocks, etc of other archaeological structures, it could be possible to define an archaeological feature.

Biases and Limitation

•Few sections of the mosaicked images produced in photoscan were blurred which made it difficult for eCognition to recognize individual rocks.
•Limitation in kite imagery causing some of the pictures to be obliqued
•eCognition could not always segment individual rocks.
•Human error in digitizing.


I would like to thank Dr. Suzanne Wechsler, Dr. Carl Lipo, Dr. Christopher Lee of California State University, Long Beach; Dr. David Burney and Lida Burney of the National Tropical Botanical Garden; Paul Nesbit, Briton Voorhees, Mike Ferris, Jacob Kovalchik, Matt Lucas, Ted Ralston, Chuck Devaney, David Hummer, John O’Conner, Samantha Hauser and Avery Sandborn.

This research was made possible by the National Science Foundation, Research Experiences for Undergradutes Program NSF Award #1005258 Geospatial Research and mapping(GRAM)


    Field Locations