Road Mapping in Lidar Images Using a Joint-Task Dense Dilated Convolutions Merging Network

Abstract

It is important, but challenging, for the forest industry to accurately map roads which are used for timber transport by trucks. In this work, we propose a Dense Dilated Convolutions Merging Network (DDCM-Net) to detect these roads in lidar images. The DDCM-Net can effectively recognize multi-scale and complex shaped roads with similar texture and colors, and also is shown to have superior performance over existing methods. To further improve its ability to accurately infer categories of roads, we propose the use of a joint-task learning strategy that utilizes two auxiliary output branches, i.e, multi-class classification and binary segmentation, joined with the main output of full-class segmentation. This pushes the network towards learning more robust representations that are expected to boost the ultimate performance of the main task. In addition, we introduce an iterative-random-weighting method to automatically weigh the joint losses for auxiliary tasks. This can avoid the difficult and expensive process of tuning the weights of each task's loss by hand. The experiments demonstrate that our proposed jointtask DDCM-Net can achieve better performance with fewer parameters and higher computational efficiency than previous state-of-the-art approaches.

Roads Detection and Parametrization in Integrated BIM-GIS Using LiDAR

“…In the final stage, road centerlines are extracted by using a Voronoi-diagram-based approach and then by removing dangle lines. Liu et al [35] specifically focused on mapping roads that are used for timber transport. In this work, a Dense Dilated Convolutions Merging Network (DDCM-Net) is used to detect roads in LiDAR data.…”

Live Extraction of Curvilinear Structures From Lidar Raw Data

“…In White et al (2010), careful manual delineation of haul roads in DTM view show comparable results to accurate field surveys. Unsupervised solutions based on DTM analysis have also been developed, using a morphological classification and stochastic geometry tools to build and assemble a graph structure (Ferraz et al, 2016), or convolution neural network architectures (Salberg et al, 2017, Liu et al, 2019. But all these tools incorporate the interpolation errors.…”

Automatic Forest Road Extraction from LiDAR Data of Mountainous Areas

“…In the literature, several methods have been developed for road extraction from LiDAR data [1,2,6,8,10,11]. Most of them rely on the shaded DTM analysis, using standard image processing tools.…”