A Study of Using Fully Convolutional Network for Treetop Detection on Remote Sensing Data

Abstract

<p><strong>Abstract.</strong> Individual tree detection and counting are critical for the forest inventory management. In almost all of these methods that based on remote sensing data, the treetop detection is the most important and essential part. However, due to the diversities of the tree attributes, such as crown size and branch distribution, it is hard to find a universal treetop detector and most of the current detectors need to be carefully designed based on the heuristic or prior knowledge. Hence, to find an efficient and versatile detector, we apply deep neural network to extract and learn the high-level semantic treetop features. In contrast to using manually labelled training data, we innovatively train the network with the pseudo ones that come from the result of the conventional non-supervised treetop detectors which may be not robust in different scenarios. In this study, we use multi-view high-resolution satellite imagery derived DSM (Digital Surface Model) and multispectral orthophoto as data and apply the top-hat by reconstruction (THR) operation to find treetops as the pseudo labels. The FCN (fully convolutional network) is adopted as a pixel-level classification network to segment the input image into treetops and non-treetops pixels. Our experiments show that the FCN based treetop detector is able to achieve a detection accuracy of 99.7<span class="thinspace"></span>% at the prairie area and 66.3<span class="thinspace"></span>% at the complicated town area which shows better performance than THR in the various scenarios. This study demonstrates that without manual labels, the FCN treetop detector can be trained by the pseudo labels that generated using the non-supervised detector and achieve better and robust results in different scenarios.</p>

Evaluation of the NEON dataset for the automatic detection of treetops and delineation of tree crowns in American and German mixed forests

“…First, the CHM of each study area is pre-processed by applying different vegetation masks (VMs) to remove irrelevant pixels. Then, the positions of the treetops are automatically detected with a method based on the work of Xiao et al (2018) that uses the morphological Top-Hat by Reconstruction (THR) operation. 35 Last, those treetop positions are utilised as seeds for delineating the tree crowns with the RG segmentation algorithm following the work of Dalponte and Coomes (2016).…”

Slope reliability analysis using Bayesian optimized convolutional neural networks

“…, 2014), human action recognition (Ji et al. , 2013), remote sensing (Xiao et al. , 2018), structural health monitoring (Cha et al.…”

“…As an advanced version of ANN, Convolutional Neural Networks (CNN) have been widely used in face recognition (Lawrence et al, 1997;, speech recognition (Abdel-Hamid et al, 2014), human action recognition (Ji et al, 2013), remote sensing (Xiao et al, 2018), structural health monitoring (Cha et al, 2017;Abdeljaber et al, 2017;Gopalakrishnan et al, 2017), and geotechnical engineering (Wang and Goh, 2021). Previous studies have demonstrated that CNN can not only effectively capture the topology of images but also properly identify the complicated relationship between soil parameters and the performance of geotechnical structures (Wang and Goh, 2021;He et al, 2021;Wang, 2022).…”

Assessment of Tree Detection Methods in Multispectral Aerial Images

“…In research similar to ours, Xiao et al [41] used a Fully Convolutional Network (FCN) [42] to detect treetops in satellite imagery. They fused the NDVI values, the DSM, and the red band into a 3-channel input.…”