Algoritma penurasan data lidar untuk penjanaan model ketinggian digital bagi kawasan tropika

Abstract

Filtering technique and the environmental factors are among the main factors, which affect Digital Elevation Model (DEM) accuracy obtained from the Light Detection and Ranging (LiDAR) data especially for steep area and covered by vegetation. Intensive research of LiDAR data filtering in tropical area is very limited and the improvement of the filtering technique using the environmental factor is very much needed. The purpose of this research is to improve the existing filtering techniques such as Progressive Morphology (PM) for DEM generation in the area covered by tropical vegetation. Initial test has been done by evaluating the filtering techniques such as PM, Adaptive Triangular Irregular Network (ATIN) and Elevation Threshold with Expand Window (ETEW) on the LiDAR data over Pekan, Pahang with slope between 0o and l0o. LiDAR DEM accuracy that was calculated based on ground reference point in mixed forest area shows that PM and ETEW filtering methods have produced minor RSME errors of A.226m and 0.192m compared to ATIN with 0.235m. Subsequent test was conducted for rubber area with slope value between 0o to 15o. The results show low RMSE error of 0.660m, 0.699m and A.717m for PM, ETEW and ATIN respectively. This shows that the slope parameter has an impact on the accuracy of the DEM, These results also demonstrate that the PM technique provides the highest accuracy. However the slope value in PM technique was based on constant value and applied to the entire LiDAR data. Compared to other filtering techniques, PM techniques provide more convenient way of improving the slope value. Improvement of PM filtering technique has been made by taking into account the actual slope value parameter and the revised method named AdapMorf algorithm. AdapMorf filtering technique was evaluated based on the slope gradient of the earth surface with the accuracy of the DEM error was evaluated for each area (i.e. mixed forest, rubber and oil palm) with slope between 0o and l5o. Three categories of assessments were carried out for each landcover and each category has a series of tests. DEM results were analyzed using RMSE error and the calculation of Type I and Type II errors. The best DEM's accuracy for AdapMorf by the types of landcover are 0.650m, 0.520m and 0.604m for mixed forests, rubber and oil palm respectively. The lowest results for Type I error are29.l7o/o,31.760/o and 35.l3Yo for rubber, mixed forest and oil palm respectively. The results for Type II error are 0.05oh,0.06% and 0.2lYs for rubber, mix forest and oil palm respectively. Due to the Type I error for AdapMorf was relatively high, the filtering technique was improved by introducing TyMof filtering technique. The tests were canied out and the results obtained show improvement in DEM's accuracy with RMSE for rubber and mixed forest are A.472m and 0.582m respectively. The Type I error for mixed forest and rubber are 28.90Yo and 19.29o/o respectively. This study shows that AdapMorf and TyMof filtering techniques were able to generate DEM with error smaller than the previous techniques for area with slope between 0o and 15". As a conclusion, AdapMorf and TyMof filtering techniques have shown that it can produce better quality of DEM for steep area and vegetated cover of tropical forest.