An automatic approach for tree species detection and profile estimation of urban street trees using deep learning and Google street view images

Choi, Kwanghun; Lim, Wontaek; Chang, Byungwoo; Jeong, Jinah; Kim, Inyoo; Park, Chan‐Ryul; Ko, Dongwook W.

doi:10.1016/j.isprsjprs.2022.06.004

Cited by 40 publications

(16 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The intersection over union (IoU, Figure 5a) is commonly used to evaluate the performance of an object detector for a specific dataset [131][132][133]. In the case of object detection and segmentation, IoU quantifies the degree of overlap between the target bounding box and the predicted bounding box [134][135][136]. The larger the overlapping area, the greater the IoU.…”

Section: Normalized Intersection Over Unionmentioning

confidence: 99%

A Novel Approach to Match Individual Trees between Aerial Photographs and Airborne LiDAR Data

Xu,

Wang,

Skidmore

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Integrating multimodal remote sensing data can optimize the mapping accuracy of individual trees. Yet, one issue that is not trivial but generally overlooked in previous studies is the spatial mismatch of individual trees between remote sensing datasets, especially in different imaging modalities. These offset errors between the same tree on different data that have been geometrically corrected can lead to substantial inaccuracies in applications. In this study, we propose a novel approach to match individual trees between aerial photographs and airborne LiDAR data. To achieve this, we first leveraged the maximum overlap of the tree crowns in a local area to determine the correct and the optimal offset vector, and then used the offset vector to rectify the mismatch on individual tree positions. Finally, we compared our proposed approach with a commonly used automatic image registration method. We used pairing rate (the percentage of correctly paired trees) and matching accuracy (the degree of overlap between the correctly paired trees) to measure the effectiveness of results. We evaluated the performance of our approach across six typical landscapes, including broadleaved forest, coniferous forest, mixed forest, roadside trees, garden trees, and parkland trees. Compared to the conventional method, the average pairing rate of individual trees for all six landscapes increased from 91.13% to 100.00% (p = 0.045, t-test), and the average matching accuracy increased from 0.692 ± 0.175 (standard deviation) to 0.861 ± 0.152 (p = 0.017, t-test). Our study demonstrates that the proposed tree-oriented matching approach significantly improves the registration accuracy of individual trees between aerial photographs and airborne LiDAR data.

show abstract

Section: Normalized Intersection Over Unionmentioning

confidence: 99%

A Novel Approach to Match Individual Trees between Aerial Photographs and Airborne LiDAR Data

Xu,

Wang,

Skidmore

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Nevertheless, they require long periods (e.g., 18 months) and a high amount of money [6]. In order 46 to alleviate the characterization and evaluation of urban trees, 47 previous works have proposed different artificial intelligence-48 based strategies; see [6], [7], [8], [9], [10], [11], [12] and 49 the references therein. It is essential to highlight that before 50 mentioned works use street view images in the electromagnetic 51 spectrum's Red, Green, and Blue (RGB) bands.…”

mentioning

confidence: 99%

“…1. In particular, 61 based on the literature review, we found that only a few works 62 are pursuing to tackle the segmentation of urban trees and the 63 computation of their dendrometric parameters [8], [10].…”

mentioning

confidence: 99%

“…The reference [8] aims to automatically determine tree 65 profile information such as the tree height, diameter at breast 66 height (DBH), and tree species. Specifically, the methodology 67 proposed starts detecting trees within a bounding box using 68 YOLOv3.…”

mentioning

confidence: 99%

“…However, large urban datasets which include tree instances have the disadvantage that they do not discriminate individual occurrences of trees [14], [15]. Moreover, as mentioned by [8], it is challenging to generate a customized semantic segmentation dataset comparable to object detection datasets [16], in terms of training samples, due to semantic segmentation requires pixel-wise labeled instances, which is an arduous task. Therefore, in the present work, we explore a strategy to improve the segmentation of urban trees by two state-of-the-art deep neural networks (i.e., DeepLabV3 and SegFormer [17], [18]).…”

mentioning

confidence: 99%

See 2 more Smart Citations

Exploring the Potential of Vegetation Indices for Urban Tree Segmentation in Street View Images

Arevalo-Ramirez¹,

Alfaro²,

Saavedra³

et al. 2023

Preprint

View full text Add to dashboard Cite

<p>Urban forests play a crucial role in the development of cities because of the urban ecosystem services they provide. Previous works have alleviated urban forest monitoring by discriminating tree species and performing tree inventories using street view images and convolutional neural networks. However, the characterization of trees from street-view images remains a challenging task. Determining tree structural parameters has been limited because of inaccurate tree segmentation caused by combined, occluded, or leaf-off trees. Therefore, the current work evaluates the potential of vegetation indices derived from red, green, blue, and synthesized near-infrared and red-edge spectral bands for urban tree segmentation. In particular, we attempt to show whether or not vegetation indices add relevant information to deep neural segmentation networks when there are low fine-tuning training samples. A conditional adversarial network generates red-edge and near-infrared images in urban environments, which retrieve an average structural similarity index of 0.86 and 0.81, respectively. Furthermore, we note that by using appropriate multispectral vegetation indices, one can boost the average intersection over the union between 5.07 % to 13.7 %. Specifically, we suggest the SegFormer segmentation network pre-trained with the CityScapes dataset and Red Edge Modified Simple Ration index for improving urban tree segmentation. However, if no multispectral data is available, the DeepLabV3 network pre-trained with the ADE20k dataset is suggested because it could achieve the best RGB outcomes average IoU value of 0.671. </p>

show abstract

Integrating planar and vertical environmental features for modelling land surface temperature based on street view images and land cover data

Wei

Guan

Zhang

et al. 2023

Building and Environment

View full text Add to dashboard Cite

An automatic approach for tree species detection and profile estimation of urban street trees using deep learning and Google street view images

Cited by 40 publications

References 32 publications

A Novel Approach to Match Individual Trees between Aerial Photographs and Airborne LiDAR Data

A Novel Approach to Match Individual Trees between Aerial Photographs and Airborne LiDAR Data

Exploring the Potential of Vegetation Indices for Urban Tree Segmentation in Street View Images

Integrating planar and vertical environmental features for modelling land surface temperature based on street view images and land cover data

Contact Info

Product

Resources

About