https://doi.org/10.59132/geo/2020/1/33-40
Povzetek
Raziskovalna naloga Vrtače Rampolj in okolice v kamniški občini je na 52. državnem srečanju mladih raziskovalcev prejela zlato priznanje v kategoriji Geografija in geologija. Med raziskovalnim delom smo odgovarjali na raziskovalna vprašanja, ali bi se dalo vrtače dovolj zanesljivo prepoznati z avtomatsko obdelavo lidarskega modela površja. Za namen raziskave smo na obravnavanem območju planot vzhodno od Kamnika na terenu popisali vse vrtače. Razvili smo novo metodo za zaznavanje vrtač, ki temelji na postopku slikovnega ujemanja. Za razliko od ostalih metod za avtomatsko zaznavanje vrtač, ki temeljijo na iskanju posamezne geometrične značilnosti terena (npr. lokalnega minimuma ali zaprtih kotanj), ta metoda temelji na iskanju območij, ki imajo obliko, podobno vrtači. Analizirali smo uspešnost predlaganega postopka s primerjavo terenskih podatkov. Kakovost zaznave ocenjujemo z deležem vrtač, ki smo jih zaznali, ter deležem zadetkov, ki v resnici niso vrtače.
Abstract
Automatic Doline Detection Using Image Matching
The research project Dolines of Rampolje and its Surroundings in the municipality of Kamnik received the Golden Award in geography at the 52nd National Meeting of Young Researchers. The scientific research question was: Can dolines be reliably detected automatically from the LiDAR elevation model? For the purpose of the research, all dolines were inventoried on the field in a welldefined plateau area east of Kamnik. A new method for the detection of dolines was established based on the Image Matching algorithm. Unlike other doline detection methods used to find a particular geometrical characteristic, e.g. local minima or enclosed sink, the proposed method is used to detect areas with a doline shape. The quality of the proposed method was evaluated based on the comparison with field observations. The quality is evaluated through the proportion of the actual dolines detected and also the proportion of false detections that are not confirmed by field observations.