The impact of user-defined parameters on the accuracy of digital terrain model

U ovom se diplomskom radu istražuje utjecaj korisničko-definiranih parametara na točnost digitalnog modela reljefa. Digitalni modeli reljefa (DMR) izrađeni su na temelju tri skupa visinskih podataka: vektorizirane izohipse i visinske točke, lidarski oblak točaka i fotogrametrijski prikupljeni uzorci. Lidarski generiran DMR poslužio je kao referentno stanje za ocjenu točnosti ostalih modela. Nad fotogrametrijski prikupljenim uzorcima testirano je deset metoda prostorne interpolacije, a najbolji rezultat ostvaruje se primjenom običnoga kriginga. Poseban naglasak postavljen je na veličinu piksela, koja je računata za svaki model napose iz nekoliko formula. Usto, razvijen je koncept hibridnog DMR-a, koji se temelji na prestrukturiranju i progušćivanju ulaznih podataka izohipsama. Rezultati ukazuju kako su, uz referentni lidarski DMR, najtočniji DMR-ovi nastali vektorizacijom HOK-a. Među fotogrametrijski izrađenim modelima, na koje je rad prvenstveno i fokusiran, hibridni DMR veličine piksela 3 m postiže bolje rezultate po svim aspektima ocjene točnosti nego DMR proizveden uobičajenim pristupom. Modeli su međusobno uspoređeni uobičajenim metodama, ali i u uspješnosti izvođenja hidroloških parametara: 1) površinskog otjecanja – primjenom razvijenog “DNK” koncepta utvrđivanja horizontalnog odmaka i 2) drenažnog bazena – izračunom volumena i trodimenzionalne površine.This thesis determines the impact of user-defined parameters on the accuracy of digital terrain model. Digital terrain models (DTMs) are made on the basis of three sets of elevation data: vectorized contour lines and elevations, LiDAR cloud points and photogrammetry samples (stereorestitution). DTM produced from LiDAR samples was used as the reference value for assessing the accuracy of the other DTMs. By using photogrammetry samples the author tested ten methods of spatial interpolation, and the best result was achieved by using the method of ordinary kriging (OK). The special emphasis is set on the pixel size, which is calculated for each model individually with several calculations. Also, the concept of a hybrid DTM was developed, which is based on the restructuring and refinement of the input data by creating geostatistical contour lines. The results indicate that the most accurate DTMs were produced from vectorized contour lines and elevations. Among the photogrammetric generated DTMs, to which the thesis is primarly focused, the hybrid model with spatial resolution of 3 m achieved much better results in all aspects of accuracy assessment, than DTM produced by the usual approach. DTMs are also compared by the success of performance of some hydrological parameters: 1) flow accumulation – where „DNA“ concept of determining horizontal discrepancy was developed and 2) drainage basin – by analysing volumes and 3D surfaces