空撮画像の同質領域判定による九州北部里山林の樹冠抽出に関する研究(平成15年度日本造園学会全国大会研究発表論文集(21))

概要

論文の詳細を見る
This paper attempts to extract data for forest crowns analysis by the determining same color regions method of image processing. For the analysis of crown size aerial photographic RGB images of broad leaved forest, also of Cryptomeria japonica and Chamaecyparis obtuse forest, of a Northern Kyushu Satoyama Rural Forest were utilized. The Image color distribution was used as an image feature. At first, the circle was calculating at every pixel less than the range of the determined standard value (h) of the color. Next, diffuseness information such as shadow areas and overlapping circles were eliminated from images. Remaining dot images were the processing data for the crown. In each of the research plot, verification data of crown width and number was ascertained by observations, and then compared with processed data. As a result, broad leave forest was determined have the best accuracy for extracting crown width at h: 40, and Cryptomeria japonica and Chamaecyparis obtuse forests were h: 15 to 20. However crown numbers were counted at less than 50% of the original images, because of the characteristics of this image processing method.
社団法人日本造園学会の論文
2003-03-31