Monitoring and modeling the process of expansion and physical development of cities by applying the combined method of neural network and neighborhood filter
Considering the ever-increasing speed of the urban population and its impact on urban growth and the physical expansion of cities and their encroachment on suitable agricultural land in the suburbs, studying, investigating, modeling and monitoring these unplanned expansions are among the important t...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | fas |
| Published: |
Assist. Prof. Dr. Roohollah Taherkhani
2022-08-01
|
| Series: | مهندسی و مدیریت ساخت |
| Subjects: | |
| Online Access: | https://www.jecm.ir/article_156401_1b5a686e420b7d32b8c5674907d763b7.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Considering the ever-increasing speed of the urban population and its impact on urban growth and the physical expansion of cities and their encroachment on suitable agricultural land in the suburbs, studying, investigating, modeling and monitoring these unplanned expansions are among the important topics of researchers. It is in different fields. For this reason, in this research, Landsat satellite images in 1996, 2006, and 2016 were used to produce the city map and the combination of neural network method with neighborhood filters was used for monitoring and modeling. In this article, only the growth of building use is considered as urban growth. Support vector machine method was used for image classification and map extraction. To model urban growth, the proposed feed-forward neural network was implemented in 2 stages; (1) to learn and determine the weights using the map of 1996 and 2006, and (2) forward to predict the merit map for 2016. In order to accurately predict the merit map Above, the neural network architecture was determined according to the lowest RMSE. Then the predicted merit map was combined with different neighborhood filters and the 2016 map was predicted. The accuracy of the method was determined in two steps. First, the accuracy of the predicted merit map was checked using the ROC method, and then, in the second step, the degree of conformity of the predicted urban map with the reference map of 2016 was obtained by using the comparison matrix and with the criteria of overall accuracy and Kappa coefficient. Finally, the presented method was used to predict the urban map of 2026. |
|---|---|
| ISSN: | 2538-1903 2538-2578 |