Abstract
Abstract. 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 % at the prairie area and 66.3 % 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.
Highlights
Forest is one of the most important land surfaces of the earth and plays an important role in the global ecosystem
Besides the pseudo labels that generated by the top-hat by reconstruction (THR) operation, we still need true labels as reference data to assess the performance of the treetop detection
We calculate the results of detection accuracy (DA), ecom and eom, as well as the precision P and F-scores and the final results can be found in table 1 and the visualized results can be found in figure 8 and figure 9
Summary
Forest is one of the most important land surfaces of the earth and plays an important role in the global ecosystem. Many works have been proposed to perform tree detection and crown delineation with remote sensing data and have shown great potential in accurately detect in individual level (Hill et al, 2017; Kathuria et al, 2016; Latifi et al, 2015). In most of these methods, the treetop detection is an essential and critical step, of which the detection accuracy is decisive for the final results. It is normally hard to find a suitable window size for the treetop detection due to the high variation of crown size, even with methods that adaptively adjust filter size (Ke and Quackenbush, 2011; Özcan et al, 2017; Santoro et al, 2013; Skurikhin et al, 2013; Song et al, 2010; Wulder et al, 2000)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
