408 Shojanoori and Shafri: Review on the Use of Remote Sensing for Urban Forest Monitoring Latif et al. (2012) emphasized that the cause for the low accuracy of tree crown delineation and tree counting is height. They used spectral infor- mation (i.e., MD and SAM) and segmentation via WorldView-2 imaging for tree crown delineation and tree counting. They demonstrated that each classified region could be considered as one tree after segmentation. The results of the segmen- tation and tree counting showed low accuracy because tree crown delineation is difficult when trees are clumped together and have similar height. Therefore, the research by Iovan et al. (2008) and Latif et al. (2012) indicated that obtaining the precise height information of trees is essen- tial in improving the accuracy of tree counting. Based on high-resolution imaging, the infor- mation is limited to 2D features and spectral, spatial, and contextual characteristics of trees. LiDAR data present trees in 3D, and the use of these data may lead to an improved accuracy of tree information, such as the height and width of crown base (Yao and Wei 2013). Yao and Wei (2013) demonstrated that the integra- tion of LiDAR and aerial imaging yielded more accurate results in tree crown delineation than when the LiDAR data were used alone (Lin and Hyyppä 2012). Approximately 81% of individual trees were detected, and incorrectly detected trees can be explained by the misclassification of small trees or buildings that block the view. Although several studies have demonstrated that the use of LiDAR data is necessary for tree crown delineation and counting, the high accuracy of the results from other studies indicated the possibility of urban tree crown delineation without the use of LiDAR data. Nevertheless, all of these researchers believed that in an urban area, the use of only the spectral information may not be feasible; thus, other information such as spatial and textural should be utilized. Moreover, processing the LiDAR data is time consuming and not as economical compared with high-resolution imaging. Thus, new high-reso- lution imaging techniques, such as the combination of WorldView-2 with an object-based technique, are recommended for use in urban tree crown delinea- tion and tree counting. However, if the cost and pro- cessing time are not important factors to consider, the use of the LiDAR data alone is sufficient in increas- ing the accuracy of urban tree crown delineation. ©2016 International Society of Arboriculture In summary, tree counting is simplified through tree crown delineation, and field surveying is no longer needed. Studies have reported that most tree crown delineation and detection processes can be conducted automatically, but the count- ing remains semi-automatic. Particular studies have shown accurate results in semi-automatic tree counting in forests (Shafri et al. 2012), but fully automatic tree counting has neither been performed in a forest nor an urban area. To sum up, this review demonstrated that studies on urban tree crown delineation and tree counting are rare and further research must be performed. CONCLUSION This paper presented the status of urban forest monitoring involving remote sensing. First, different remote sensors used to generate urban vegetation maps were evaluated. Second, various classification methods used to extract urban for- est information and distinguish urban tree species were assessed. Third, different methods for tree crown delineation and tree counting were dis- cussed. This paper considered the most significant problems and mentioned the solution based on remote-sensing methods through related studies. Remote-sensing imaging can detect urban for- ests, but different sensors have their own limita- tions. For instance, the use of hyperspectral data is optimal in extracting and distinguishing the urban forest based on spectral information, but high volume data, availability, and cost are limi- tations. The other data, which can detect urban features even in shadow and at night, is LiDAR data. However, the use of this data leads to mis- classification and is time-consuming because it involves the conversion of point clouds into raster layers. Thus, other high-resolution imag- ing techniques, such as WorldView-2, are utilized. Urban areas are complex environments, and the limitation of spectral information for mul- tispectral imaging, particularly for distinguish- ing tree species, has compelled researchers to utilize other urban vegetation information during classification; such information includes spa- tial, textural, and color. Hence, the object-based approach is more applicable than traditional pixel-based classifications, such as MLC and MD, because the former can combine various data.
November 2016
| Title Name |
Pages |
Delete |
Url |
| Empty |
Search Text Block
Page #page_num
#doc_title
Hi $receivername|$receiveremail,
$sendername|$senderemail wrote these comments for you:
$message
$sendername|$senderemail would like for you to view the following digital edition.
Please click on the page below to be directed to the digital edition:
$thumbnail$pagenum
$link$pagenum
Your form submission was a success. You will be contacted by Washington Gas with follow-up information regarding your request.
This process might take longer please wait
