406 Shojanoori and Shafri: Review on the Use of Remote Sensing for Urban Forest Monitoring leads to a high spectral similarity between veg- etation in urban areas. Furthermore, urban areas contain numerous pollutants that can change atmospheric conditions and affect spectral reflec- tance (Iovan et al. 2014). As a result, spectral sig- natures in multispectral imaging are insufficient to distinguish urban tree species. Thus, other characteristics of a tree species, such as spatial information, texture, and color, should be utilized to improve the classification of urban tree species. The use of object-based classification is recom- mended to overcome the spectral limitation of multispectral imagery. Shouse et al. (2013) com- pared two classification methods (i.e., pixel- and object-based classification) in two types of mul- tispectral imaging techniques (i.e., aerial and LandSat TM 5) to detect a species called bush honeysuckle (i.e., Lonicera maackii). The results showed that the object-based approach presents higher accuracy than the pixel-based approach, and HSR imaging demonstrates high accuracy [aerial (HSR): 94.2% / Landsat (MSR): 74.6%]. Textural information is an effective type of information that can be used to distinguish tree species. Iovan et al. (2008; 2014) used HSR data and SVM to distinguish urban tree species (Plat- anus, Sophora, Tilia, Celtis, Pinus, and Cupres- sus). As spectral information was inappropriate to be used independently, textural information, which involves information regarding the spa- tial and physical arrangement of objects, was utilized (Tso and Mather 2001). The results dem- onstrated that both methods of textural mea- surements (i.e., first- and second-order GLCM) could detect urban tree species and distin- guish deciduous trees from coniferous species. LiDAR data are optimal sources of informa- tion on texture or height. LiDAR data have been used in many studies on urban forest species (Voss and Sugumaran 2008; Tooke et al. 2009; Hofle et al. 2012; Zhang and Qiu 2012; Forzieri et al. 2013; Tigges et al. 2013). A robust technique for classification or segmentation is needed when the amount of information increases. Hofle et al. (2012) showed that intelligence algorithms, such as the ANN, are suitable for LiDAR informa- tion analysis. He applied two methods according to the object based-approach (ANN and DT) to detect six tree species, namely, Fagus sylvatica, ©2016 International Society of Arboriculture Acer platanoides, Platanus acerifolia, Tilia cordata, platyphyllos, and Aesculus hippocastanum. The result showed that ANN presents a higher accu- racy of 95% than DT with an overall accuracy of 72%. Spatial, textural, shape, or height informa- tion from LiDAR data can also be used to detect tree species. Zhu et al. (2012) showed that the spectral characteristics from LiDAR data are applicable to distinguishing real leaves from fake ones. Despite these results, trees oſten have approximately the same height and shape, and the high density of tree species may lead to the mis- classification of tree species or small trees may be overlooked (Iovan et al. 2008; Latif et al. 2012). The challenges of high-resolution imaging for the detection of urban tree species were highlighted when the new HSR imaging tech- nique called WorldView-2 was launched. Pu and Landry (2012) attempted to perform segmen- tation via two methods (i.e., LDA and regres- sion trees) to detect seven tree species (i.e., sand live oak, laurel oak, live oak, pine, palm, camphor, and magnolia) and demonstrated that the four new bands of WorldView-2 imag- ing can improve the accuracy by about 16% to 18% (compared with IKONOS imaging). This review showed that given the complexity of urban areas and spectral similarity between tree species, high-resolution imaging via pixel-based methods is insufficient to discriminate urban tree species; hence, ancillary data such as DEM, spa- tial information, texture, and color, should be uti- lized. Although the use of LiDAR data, as ancillary data, presents high classification accuracy, higher classification accuracy cannot be achieved when these data are used separately. Thus, two imag- ing techniques (HSR and LiDAR) are proposed to obtain high classification accuracy, but these methods are not cost effective. By contrast, the object-based method was used to distinguish spe- cies by using information about urban tree species. In this regard, different high-resolution satellite imaging techniques were used, and WorldView-2 showed the highest accuracy over other high- resolution satellite imaging techniques. Accord- ingly, for urban tree species detection, the object- based technique should be improved and integrated with WorldView-2 or with new high-resolution imaging techniques, such as WorldView-3.
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
