Extraction of linear structures from digital terrain models using deep learning

Abstract

Abstract. This paper explores the role deep convolutional neural networks play in automated extraction of linear structures using semantic segmentation techniques in Digital Terrain Models (DTMs). DTM is a regularly gridded raster created from laser scanning point clouds and represents elevations of the bare earth surface with respect to a reference. Recent advances in Deep Learning (DL) have made it possible to explore the use of semantic segmentation for detection of terrain structures in DTMs. This research examines two novel and practical deep convolutional neural network architectures i.e. an encoder-decoder network named as SegNet and the recent state-of-the-art high-resolution network (HRNet). This paper initially focuses on the pixel-wise binary classification in order to validate the applicability of the proposed approaches. The networks are trained to distinguish between points belonging to linear structures and those belonging to background. In the second step, multi-class segmentation is carried out on the same DTM dataset. The model is trained to not only detect a linear feature, but also to categorize it as one of the classes: hollow ways, roads, forest paths, historical paths, and streams. Results of the experiment in addition to the quantitative and qualitative analysis show the applicability of deep neural networks for detection of terrain structures in DTMs. From the deep learning models utilized, HRNet gives better results.

Identification of historical trackways in forests using contextual geospatial analyses

“…Although machine learning methods such as CarcassonNet (Verschoof‐van der Vaart & Landauer, 2021) or network friction route zones (van Lanen et al, 2015; Vletter & van Lanen, 2018) have recently been increasingly preferred across different geographic scales, the efficiency of the algorithm is determined primarily by the experts who identified and extracted features of the subjects under investigation (Satari et al, 2021). Even the use of modern software may result in data misinterpretation (Seifried & Gardner, 2019).…”

Flat roof classification and leaks detections by Deep Learning

“…While RGB images are commonly used as the primary dataset, and most backbone models require RGB images, digital elevation models (DEMs) can also be utilized. Multi Directional Hillshade based on DEM data with three bands with 8 bits similar to RGB are suitable for Deep learning [24]. Another option is to leverage elevation data directly from the DEM, along with derived features such as slope (ranging from 0° to 90°) and terrain curvature [25].…”

Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMs

“…With the increasing success in application of deep learning techniques in many fields, researchers in archaeology are also using deep learning in their tasks. (Caspari and Crespo, 2019;Guyot et al, 2018), archaeological monument segmentation (Kazimi et al, 2019), and extraction of terrain structures (Satari et al, 2021).…”