Using LiDAR in combination with aerial photographs to model and discriminate green small landscape elements

Abstract

Small landscape elements are an essential part of the Dutch landscape. They shape the identity of a region and landscapes in general. Small landscape elements prevent erosion, purify water, form a habitat for many animals and birds and contribute to the recreational attractiveness of landscapes. Because of industrialisation and the increasing scale of agriculture small landscape elements are under stress. The Dutch government is aware of the role that small landscape elements play in the quality of our landscapes. The government and organisations concerned with landscape management and policy support research and grant subsidies that help in the preservation of these landscape elements. It is however still difficult to monitor the state of different types of small landscape elements and the changes that are appearing. This is because there is no objective quantitative dataset with small landscape elements of the Netherlands available on a national scale. Therefore, the goal of this research is to make a model that can detect green small landscape elements. Geographical Information Systems are used to make this model and several remote sensing techniques are combined to get the desired result. To discriminate small landscape elements segmentation techniques are used on LiDAR data. In the research two areas are used: a training area (Chaam) and a validation area (Wageningen). In both areas the accuracy of the model is tested by adding a false colour image to the LiDAR data to see if this improves the model. This research has shown that LiDAR is a very promising technique for classifying green small landscape elements. Adding a false colour image to the LiDAR data is especially useful in areas (such as the Wageningen area) where there are also man-made objects. There is however still work to be done to better detect tree rows and lanes. Future work can benefit from this model and improve it (for example by adding tree crown detection by using a region growing algorithm) so that all small landscape elements can be detected.