Back to results list
Please use this identifier to cite or link to this item:
|Title:||An integrated spatial metrics and texture analysis method for classification of urban land cover and land use||Authors:||Ruiz Hernandez, Ivan Elias||Advisors:||Shi, Wenzhong (LSGI)||Keywords:||Land cover
Land use mapping
|Issue Date:||2018||Publisher:||The Hong Kong Polytechnic University||Abstract:||Since 2007, half of the world's inhabitants have been living in urban areas and by the year 2050, it is projected that the urban population will surpass 60% of the total population. Rapid urban growth is causing various urban planning related issues. Accurate and detailed spatial data on the natural and built environment is essential to solve, evaluate, monitor and support the urban development of cities. There is a lack of proper mapping approaches and spatial models, mainly due to the heterogeneity of factors involved in the complex urban dynamics and the rapid growth of cities. The identification of urban land use and land cover play a significant part in decision making for construction and the development of infrastructure. The traditional way of extracting land use and land cover is through the classification of satellite images with the help of statistical algorithms. These algorithms group similar values using the spectral information of the image and is performed using a pixel or object-based approach. This is appropriate for land cover but for land use, spatial data is necessary but rarely used. Recently, with the improvement of remote sensing technologies, higher spatial resolutions and newer algorithms permit for hybrid workflows. These workflows allow for the integration of data from different sources, which improves the overall accuracy. However, the heterogeneity of urban landscapes makes the classification challenging. Moreover, the processes to update the government databases are based in visual interpretation and on-screen digitizing using aerial photography, field survey, and census data. This convoluted process produces additional challenges as it is not only time consuming but is also costly.
To control and analyze urban growth, new and better methods for urban land use mapping are needed. This dissertation proposes a new method, which integrates spatial metrics and texture analysis in an object-based image analysis classification. The classification workflow follows an object-based approach using a high-resolution satellite image. To be able to process the land cover classification the machine learning algorithm, Random Forests, was employed. Following which, the texture analysis values were extracted from the satellite image. Then, to produce the spatial component for land use, landscape metrics were created from the land cover classification. Finally, the classification of land use, which integrates landscape metrics and texture values with the spectral data in an object-based image analysis, was also performed using Random Forest. The advantages of the proposed method are: 1) the use of Random Forests is computationally much more efficient, which allows for dealing with a large number of features and faster classifications, 2) the randomness factor of the algorithm decreases the uncertainty between classes, 3) the co-occurrence measures from the texture analysis help to improve the final overall accuracy, 4) the landscape indices can distinguish between different types of residential areas and land uses, 5) it achieves better performance when compared to different classification approaches and 6) it works without the use of any survey or census data. The efficiency and accuracy of the proposed method were validated with a case study experiment. The final classification was tested using a 10-fold cross-validation scheme, achieving an overall accuracy of 92.3% and a kappa coefficient of 0.896. This method produced an accurate model of urban land use, without the use of any ancillary data, and suggests that the combined use of landscape metrics and texture is promising. Urban land-use is crucial to urban planning and requires a detailed understanding of the city's characteristics in an efficient, accurate and in a timely manner. The maps produced from the method proposed in this dissertation can provide the land use data needed by urban planners for effective planning. This method is expected to be an essential asset for urban planners to solve the different issues that arise.
|Description:||xii, 230 pages : color illustrations
PolyU Library Call No.: [THS] LG51 .H577P LSGI 2018 Ruiz Hernandez
|URI:||http://hdl.handle.net/10397/79542||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|991022164555303411_link.htm||For PolyU Users||167 B||HTML||View/Open|
|991022164555303411_pira.pdf||For All Users (Non-printable)||34.13 MB||Adobe PDF||View/Open|
Citations as of Dec 10, 2018
Citations as of Dec 10, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.