Land Cover Segmentation of Airborne LiDAR Data Using Stochastic Atrous Network

Abstract

Inspired by the success of deep learning techniques in dense-label prediction and the increasing availability of high precision airborne light detection and ranging (LiDAR) data, we present a research process that compares a collection of well-proven semantic segmentation architectures based on the deep learning approach. Our investigation concludes with the proposition of some novel deep learning architectures for generating detailed land resource maps by employing a semantic segmentation approach. The contribution of our work is threefold.(1) First, we implement the multiclass version of the intersection-over-union (IoU) loss function that contributes to handling highly imbalanced datasets and preventing overfitting. (2) Thereafter, we propose a novel deep learning architecture integrating the deep atrous network architecture with the stochastic depth approach for speeding up the learning process, and impose a regularization effect. (3) Finally, we introduce an early fusion deep layer that combines image-based and LiDAR-derived features. In a benchmark study carried out using the Follo 2014 LiDAR data and the NIBIO AR5 land resources dataset, we compare our proposals to other deep learning architectures. A quantitative comparison shows that our best proposal provides more than 5% relative improvement in terms of mean intersection-over-union over the atrous network, providing a basis for a more frequent and improved use of LiDAR data for automatic land cover segmentation.

Using the U‐net convolutional network to map forest types and disturbance in the Atlantic rainforest with very high resolution images

“…These deep learning algorithms are used for land cover classification (Arief et al. ; Sun et al. ), scene classification (Maggiori et al.…”

“…For example, in the case of image segmentation, prior information is only given by labeled masks of the objects to recognize in the training images. These deep learning algorithms are used for land cover classification (Arief et al 2018;Sun et al 2018), scene classification (Maggiori et al 2017;Wang et al 2017;Liu et al 2018) and object extraction (Xu et al 2018). One particular type of network used for object extraction, the Unet network, is highly promising as it has been shown to outperform all traditional classification methods (Ronneberger et al 2015;Huang et al 2018).…”

A Point-Wise LiDAR and Image Multimodal Fusion Network (PMNet) for Aerial Point Cloud 3D Semantic Segmentation

“…In order to address this fundamental query, this research focuses on the 3D point cloud segmentation with various fusion and non-fusion approaches and evaluated the performance. If the data representation similar or transformed into a similar representation as in the case of [10,13], multimodality fusion can be carried out in numerous ways. Nonetheless, both data representation and range of values are different in LiDAR point cloud and images, and hence, the fusion approach has to respect the characteristics of both modality.…”

“…Several previous studies [9][10][11] have been proposed multimodal aerial data usage for segmentation, where, the point cloud is directly interpolated to the resolution of aerial images and further segmented by considering it as a 2.5D data [12,13]. Pan et al [12] uses fully connected layers for fusing CNN derived multimodal features, while improved method [13] uses an end-to-end multi-level fusion using Fully Convolutional Network (FCN).…”

Convolutional Neural Networks for On-Board Cloud Screening

“…Well-known convolutional neural networks (CNNs) have advanced the state-of-the-art in many computer vision tasks such as image classification and semantic segmentation [16][17][18][19][20][21]. Not an exception to such trend, remote sensing has been also enjoying the benefits of deep learning algorithms in achieving state-of-the-art performance in many tasks such as land-use classification and segmentation [22][23][24][25][26][27]. CNNs usually consist of thousands of filters with millions of learnable parameters which are trained to detect semantic representations useful for a particular task and over a large amount of annotated images.…”