Smart Urban Cadastral Map Enrichment—A Machine Learning Method

Round 1

Reviewer 1 Report

Thank you for submitting your interesting work. The manuscript is well written and sections are well connected. My overall critique is attached as file.

Comments for author File: Comments.pdf

Moderate editing required

Author Response

Smart Urban Cadastral Map Enrichment- A Machine Learning Method

Manuscript ID: ijgi-2799545

Reply to comments:

Attention: Prof. Wolfgang Kainz,                                                                       Date: Feb. 1, 2024

Editor-in-Chief, ISPRS International Journal of Geo Information,

Dear Prof. Wolfgang Kainz,

Thank you for your kind valuable comments and we appreciate all reviewers’ valuable comments and suggestions which have significantly improved the quality of the paper. We have addressed and incorporated all the reviewers’ suggested point by point and have revised the paper accordingly.

The revised parts in the paper have been presented in red color. In this answer letter, the respected reviewers’ questions, comments and recommendations are presented in black color and numbered as Qi where i is the number of questions, comments and recommendations number and the answers are mentioned in green color with Ai, where i is the of number of our answers. Please see below the details of the revision.

With bets wishes

Prof. Mahmoud Reza Delavar,

University of Tehran

On behalf of all the co-authors.

Reviewer 1:

Line 33-34: The statement, “Cities are becoming more complex day to day and their measuring, managing, monitoring, mapping and modeling the urban changes….”

Q1: Due to which reasons, you think cities are becoming more complex?

A1: Extra explanation is added and highlighted in the new revision. “Cities are becoming more complex day to day due to increments in population, urbanization and pace of construction and changes in the cities”.

Q2: How the complexities make challenging measuring, managing, monitoring, mapping and modeling the urban changes? Please elaborate

A2: Extra explanation is added and highlighted in the new revision. “Vast extent, high density and fast changes in cities and urban areas cause measuring, managing, monitoring, mapping and modeling the urban changes to become more challenging and time-consuming for the concerned organizations and researchers.”

 Line 40: “Urban cadastral maps as a part of LAS…..”

Q3: What are other components of a LAS. Please describe for readers’ clarity.

A3: Extra explanation is added and highlighted in the new revision. “Land information is required for different LAS components which are, land development, land use, land value and land tenure [5]. Urban cadastral maps as a part of land information are large-scale maps that contain dimensions, shapes, spatial relationships and different attributes of information about land parcels [6].”

Q4: Line 46: The statement, “ The urban cadastral maps are enriched by documented and geometric changes, new information and data or eliminating incorrect data”. It is suggested to rephrase it.

A4: It is rephrased to “The urban cadastral maps are enriched with the help of legal documents for geometric changes, incorporating new information and data or the removal of inaccurate data”.

Q5: Line 49: The statement, “Since map enrichment is a parcel-based time-consuming process…..”

Please elaborate the following

• Why enrichment of maps especially parcel-based is time consuming?
• How many steps are involved in enrichment of parcel-based maps? And
• by machine learning algorithms which steps can be automated?

A6: Extra explanations are added and highlighted in the new revision. “Map enrichment is a parcel-based time-consuming process due to a large number of parcels and information in urban areas and needs visual parcel-to-parcel checking and a high level of expertise in the different steps of enrichment including matching, change detection and base map enrichment. Hence, detecting changes in the map with more up-to-date maps from other sources automatically by machine learning algorithms can reduce the time and cost of map enrichment”

Q7: Line 52-54: “Machine learning methods are good methods for large-scale datasets such as urban cadastral maps and supervised ensemble learning methods such as random forest have computational advantages for large-scale spatial data analyses”. It is suggested to rephrase.

A7: It is rephrased to “Machine learning methods such as random forest are good methods with computational advantages for large-scale urban cadastral maps and spatial data analyses.”

Q8: Line 59-62: “Land use is important information for urban cadastral maps that affects urban management and planning and influences tax, land conflicts and reduces damages from natural and man-made disasters. Hence, land use changes in urban maps have been investigated in many researches” , seems redundant as the manuscript focuses on geometry of parcels only.

A8: It is removed from the text in the new revision.

Q9: Line 74: The statement, “Two urban maps have been employed in this research”.

• One would expect to know about characteristics of the utilized maps such as map scale, projection system, and year. You have mentioned these but at a very later stage (Line 336 and Line 344). It is therefore suggested to describe these characteristics in Line 74.

A10: Extra explanations are added and highlighted in the new revision. “A base map (Cadaster Department map, a maps at 1:2000 scale in UTM (Zone 39N) projection systems, produced in 2002) which should be enriched and a more up-to-date urban map (Tehran Municipality map, a map at 1:1000 scale in UTM (Zone 39N) projection systems, produced in 2014) that is used to enrich the base map”.

Q10: Line 77: “….the focus of this research has been on the geometry of the parcels”.

A10: Extra explanations are added and highlighted in the new revision. “Parcels in each map are target features in this research and the focus of this research has been only on enriching the geometry of the parcels. Due to limitations in data access, other information for each parcel such as registration, legal, and descriptive information have not been investigated in this research”.

Q11: Line 79: “In this research, after preprocessing the data,………………..”. Please elaborate on preprocessing of data

A11: Extra explanations are added and highlighted in the new revision. “In this research, after preprocessing the data including format and scale homogenization and topology check, the matching algorithm has been employed to find the corresponding parcels for each parcel in each of the maps. …………….”

Q12: Line 81-83: The statement, “Then, different parameters have been calculated for each parcel in each of the maps to be used in Support Vector Machine (SVM) and Random Forest (RF) machine learning algorithms for detecting changes on the base map intelligently”.

• How many and which parameters of each map? Please add details

A12: In this part of “The proposed methodology” we only explained an overview of the methodology and further in the sub-sections of this section we explained in detail. In this case, we explained parameters in section 2-2. In addition, extra explanations are added and highlighted in the new revision. “Then, 38 different geometric, topologic and statistical parameters have been calculated …….”

Q13: Line 84: Apart from hyperparameters optimization, what other techniques  can be used to increase the accuracy of change detection. And what is the rationale for using hyperparameters optimization technique?

A13: Extra explanations are added and highlighted in the new revision. “There are different methods to increase accuracy of models including data addition, dimension reduction, regularization, missing values and outlier’s management, feature engineering, feature selection, ensemble learning and hyperparameters optimization. Missing values and outlier’s management, feature engineering, feature selection, ensemble learning (using random forest algorithms) and hyperparameters optimization have been employed in this research in the models training steps. In training the models, feature selection and hyperparameters optimization have been used to increase the accuracy of change detection.”

Q14: Line 107-108: The statement, “In this research, to identify……………gravity centers of parcels in the base map have been calculated”. What are gravity centers? Are these center points of parcels? Please elaborate how the gravity centers of parcels have been calculated?

A14: Extra explanations are added and highlighted in the new revision “The gravity center is the center of mass of a parcel that may fall inside or outside the parcel. The coordinate of gravity center point (C) of a non-self-intersecting parcel is calculated by Equations (1) and (2) …”.

Q15: Line 354-355: The statement, “Hence, the Tehran municipality map was generalized from (1:1000) to (1:2000) scale” leads to the following questions:

• Which method was used for generalization?
• Cartographic generalization is mainly realized through operators, such as aggregation, classification, enhancement and exaggeration.
• Were the parcels produced by generalization in harmony with map scale?

Please give reply of above questions.

A15: Extra explanations are added and highlighted in the new revision. “Hence, the Tehran municipality map was generalized from 1:1000 to 1:2000 scale by polygon simplification and the Douglas-Peucker algorithm with 0.1 meter tolerance. This method has preserved polygons’ shape and removed vertices with more than 0.1 meter perpendicular distance to the new line added after the vertices removal”.

Q16: General Comments: The methodology for change detection is solely dependent on matching centers of parcels. This means, if centers of parcels do not match in the corresponding map then there is no change which seems to me illogical and limits scope of the work. Please rectify/elaborate on it.

A16: As pointed out in the paper, the matching method has been implemented two-way, once from the base map to the second map and another time from the second map to the base map to ensure that all types of matching such as splitted, merged, deleted and added parcels were covered by the algorithm. Since, the target features are parcels and they have big areas, calculating their gravity center points and moving them to the other map caused center points to be located in the corresponding parcels despite small shifts and rotates in the maps. In addition, the validation shows that this method has a very good performance with 0.952 accuracy and also correct matching for complex parcels. Finally, if a parcel matches incorrectly, it will be detected in the modeling.

Q17: Line 563 566: The statement, “ After modeling all the parcels and detecting the changes in the whole area, the parcels from other maps are placed in the base map by affine transformation algorithm, and the topologic errors are removed so that the base map is intelligently and automatically enriched is confusing as one has to calculate affine transformation parameters . Please elaborate on it.

A17: Extra explanations are added and highlighted in the new revision. “After modeling all the parcels and detecting the changes in the whole area, the parcels from other maps that correspond to the changed parcels in the base map are placed in the base map by affine transformation algorithm after calculating affine parameters from unchanged neighbor parcels for each changed parcels, then the topologic errors are removed so that the base map is intelligently and automatically enriched.”.

Q18: It is suggested to add policy and practical implications of the work.

A19: Extra explanations are added and highlighted in the new revision. “In this research, unlike previous studies [6,9,11], the changes in the base map have been identified by machine learning algorithms in an intelligent and automatic method, which increases the speed and accuracy of the urban map enrichment process. Moreover, relocating changed parcels automatically and updating them with the help of a more up-to-date map reduces human errors and the time of enriching urban maps. In previous studies [6,9,11], enrichment has often been done manually, visually and case by case, which is very time-consuming compared to the proposed model. The presented framework plays a very important role in reducing the cost and time of organizations in preparing urban maps or enriching urban cadastral maps by quick and accurate use of available maps in other organizations. This framework solves some of the needs of organizations and lets them make more accurate and informed decisions.

Since the juridical map and registered parcels are fixed properties and obtained from the special administrative procedures, these parcels cannot be changed and replaced by other parcels without legal processing. In this research, the base map from the Cadaster Department was produced by NCC and had no registered parcel. This base map is usually used beside another dataset for registered parcels to handle new requests for property registration. Hence, enriching this base map reduces the Cadastral Department process for registration. On the other hand, other organizations such as utilities and infrastructure organizations can effectively use the proposed framework to enrich their maps for better decision-making and urban management.”

Q20: It is suggested to add limitations of the proposed framework. Moreover, the method should be applied to larger and different areas to check its completeness.

A20: Extra explanations are added and highlighted in the new revision. “Data access limitation was one of the main issues of this research. More information about parcels such as owner name, and land use in each map may help to match the parcels better and cover other parts of map enrichment such as semantics. The parcels with holes or parcels with their gravity centers located outside of their corresponding parcels in the other map may cause errors in the matching process. There were not a number of these types of parcels in the study area , however, these parcels can be detected by a point in polygon analysis and correct the matching process before the modeling. Since accuracy and selected features in both of the models were close, the ensemble methods for integrating the models output were not very useful. Other models that cover other parts of feature space can be integrated with these models by ensemble algorithms such as voting and stacking, to cover uncertainty”.

Unfortunately, due to data access limitations in the Cadaster department and Tehran municipality, we could not access other districts and larger areas. In addition, Zone 3 of District 6 of Tehran is one of the dense areas in Tehran which has changed sigificantly in recent years. Hence, this area was selected to train the models for different and large types of changes in the maps. Moreover, each municipality zone has its department and managing group and also its distinctive urban fabric. Hence, these map enrichments are usually requested in zonal areas. Finally, as mentioned in the framework we can retrain the models for new and larger areas and also we can split the area into smaller zones and then use trained models for modeling.

Author Response File: Author Response.pdf

Reviewer 2 Report

The purpose of the paper is to propose a smart framework to enrich a cadastral base map using a more up-to-date map automatically by machine learning algorithms. The authors have tested RF and SVM ML models combined with Genetic algorithms to finetune the hyperparameters. The implementation and validation of the models were done according to the established standards.

Unfortunately, the authors have not provided a state of the art in terms of proposed approaches in the literature that address similar research questions. Indeed, many scientific papers have investigated the use of ML and DL for aligning and updating cadaster maps and data fusion, such as :

·       Fetai, B., Grigillo, D., & Lisec, A. (2022). Revising cadastral data on land boundaries using deep learning in image-based mapping. ISPRS International Journal of Geo-Information, 11(5), 298.

·       Girard, N., Charpiat, G., & Tarabalka, Y. (2019). Aligning and updating cadaster maps with aerial images by multi-task, multi-resolution deep learning. In Computer Vision–ACCV 2018: 14th Asian Conference on Computer Vision, Perth, Australia, December 2–6, 2018, Revised Selected Papers, Part V 14 (pp. 675-690). Springer International Publishing.

·       Nyandwi, E., Koeva, M., Kohli, D., & Bennett, R. (2019). Comparing human versus machine-driven cadastral boundary feature extraction. Remote sensing, 11(14), 1662.

·       Wang, S., & Li, W. (2021). GeoAI in terrain analysis: Enabling multi-source deep learning and data fusion for natural feature detection. Computers, Environment and Urban Systems, 90, 101715.

In addition, the authors have not highlighted their main contributions compared to the existing approaches and what are the scientific gaps that they would like to address.

The quality of English is good. Minor editing and proofreading are required.

Author Response

Smart Urban Cadastral Map Enrichment- A Machine Learning Method

Manuscript ID: ijgi-2799545

Reply to comments:

Attention: Prof. Wolfgang Kainz,                                                                       Date: Feb. 1, 2024

Editor-in-Chief, ISPRS International Journal of Geo Information,

Dear Prof. Wolfgang Kainz,

Thank you for your kind valuable comments and we appreciate all reviewers’ valuable comments and suggestions which have significantly improved the quality of the paper. We have addressed and incorporated all the reviewers’ suggested point by point and have revised the paper accordingly.

The revised parts in the paper have been presented in red color. In this answer letter, the respected reviewers’ questions, comments and recommendations are presented in black color and numbered as Qi where i is the number of questions, comments and recommendations number and the answers are mentioned in green color with Ai, where i is the of number of our answers. Please see below the details of the revision.

With bets wishes

Prof. Mahmoud Reza Delavar,

University of Tehran

On behalf of all the co-authors.

Dear Respected Reviewer 2,

Thank you very much for your valuable comments which greatly improved the quality of the paper. The point by point response to youe comments are as follows:

Reviewer 2:

Q21: Unfortunately, the authors have not provided a state of the art in terms of proposed approaches in the literature that address similar research questions. Indeed, many scientific papers have investigated the use of ML and DL for aligning and updating cadaster maps and data fusion

A21: Some more recent and state-of-the-art research has been added to the new revision. Since most of their methodologies were raster-based, our contribution was to create a methodology to enrich maps in vast areas and in vector space.

Extra explanations are added and highlighted in the new revision. “Artificial intelligence methods including machine learning and deep learning have been recently used very much for map enrichment, updating, object, boundary and change detection and spatial data fusion [14,15,18–20]. In [18], a combination of data-level fusion and feature-level fusion has been employed for natural object detection by multi-source geospatial data and deep learning. Deep learning methods have been implemented to align and update cadaster maps with satellite images [19], detect visible land boundary automatically with aerial images and use to revise existing cadastral maps [20]. In [21] a comparison between manual approaches and machine learning algorithms has been done for extracting visible cadaster boundaries from satellite images in rural and urban areas and the authors have shown less cost and time and more accuracy of the machine learning algorithm against the manual approaches.”

Q22: In addition, the authors have not highlighted their main contributions compared to the existing approaches and what are the scientific gaps that they would like to address.

A22: Extra explanations are added and highlighted in the new revision in the introduction and discussion sections.

 “Most previous research has been focused on aerial or satellite images to enrich base maps. Most of them have been employed in small areas or enriched the map features case by case and there is less attention to enriching base maps with more recent maps in a vector space in a vast, dense and complex urban area. In addition, automatic change detection and map modification with an acceptable accuracy have been less investigated. Hence, the contribution of this research is to propose an intelligent framework to enrich cadastral parcels in the base map with a more up-to-date map automatically in urban areas.”

“In this research, unlike previous studies [6,9,11], the changes in the base map have been identified by machine learning algorithms in an intelligent and automatic method, which increases the speed and accuracy of the urban map enrichment process. Moreover, relocating changed parcels automatically and updating them with the help of a more up-to-date map reduces human errors and the time of enriching urban maps. In previous studies [6,9,11], enrichment has often been done manually, visually and case by case, which is very time-consuming compared to the proposed model. The presented framework plays a very important role in reducing the cost and time of organizations in preparing urban maps or enriching urban cadastral maps by quick and accurate use of available maps in other organizations. This framework solves some of the needs of organizations and lets them make more accurate and informed decisions.”

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper “Smart Urban Cadastral Map Enrichment - A Machine Learning Method” proposes enrichment of a cadastral base map using a more up-to-date map automatically by machine learning algorithms.

Unfortunately, despite the correct methodology, the manuscript has wrong approach. The purpose of the research is wrongly set.

Cadastral map and Urban map are two completely different products. Polygons on the cadastral map (cadastral parcel) are property boundaries obtained through a special administrative procedure with the participation of the holders of rights, restrictions and responsibilities. The cadastral parcel shows the spatial extension of the land tenure that is registered in the registry.

Polygons on the urban map are a technical category that does not have an administrative component. They only give the topographic features of the land. They may be similar to cadastral parcels, but we do not have confirmation of the land tenure holder for that polygon.

In your research, for example, you propose merging two cadastral parcels. They may have the same land use and topographically look like one. But, from the information you had, we don't know if they have the same owner. Such merging of cadastral parcels causes unlawful mistakes in the cadastre concerning land tenure integrity.

I suggest that you apply the methodology to another example.

Author Response

Smart Urban Cadastral Map Enrichment- A Machine Learning Method

Manuscript ID: ijgi-2799545

Reply to comments:

Attention: Prof. Wolfgang Kainz,                                                                       Date: Feb. 1, 2024

Editor-in-Chief, ISPRS International Journal of Geo Information,

Dear Prof. Wolfgang Kainz,

Thank you for your kind valuable comments and we appreciate all reviewers’ valuable comments and suggestions which have significantly improved the quality of the paper. We have addressed and incorporated all the reviewers’ suggested point by point and have revised the paper accordingly.

The revised parts in the paper have been presented in red color. In this answer letter, the respected reviewers’ questions, comments and recommendations are presented in black color and numbered as Qi where i is the number of questions, comments and recommendations number and the answers are mentioned in green color with Ai, where i is the of number of our answers. Please see below the details of the revision.

With bets wishes

Prof. Mahmoud Reza Delavar,

University of Tehran

On behalf of all the co-authors.

Dear respected reviewer 3,

Thank you for your valuable comments which have greatly improved the quality of the paper. Please find below the point-by-point answer to your comments.

Reviewer 3:

Q23: Unfortunately, despite the correct methodology, the manuscript has wrong approach. The purpose of the research is wrongly set.

Cadastral map and Urban map are two completely different products. Polygons on the cadastral map (cadastral parcel) are property boundaries obtained through a special administrative procedure with the participation of the holders of rights, restrictions and responsibilities. The cadastral parcel shows the spatial extension of the land tenure that is registered in the registry.

Polygons on the urban map are a technical category that does not have an administrative component. They only give the topographic features of the land. They may be similar to cadastral parcels, but we do not have confirmation of the land tenure holder for that polygon.

In your research, for example, you propose merging two cadastral parcels. They may have the same land use and topographically look like one. But, from the information you had, we don't know if they have the same owner. Such merging of cadastral parcels causes unlawful mistakes in the cadastre concerning land tenure integrity.

I suggest that you apply the methodology to another example.

A23: As mentioned in the paper, the Cadaster Department map (the base map) is a base map produced by Iranian National Cartographic Center (NCC) which was updated by other organizations’ maps. This map did not contain any registered parcels and there is another dataset for registered parcels in Cadaster Department. The enrichment process does not change the registered dataset and the enriched map is used to reduce procedures, time and cost of registration. Hence, if parcels are merged or splitted and it is illegal the expert in the Cadaster Department can realize and make a correct decision for registration. In addition, it has been mentioned that in this research we only focused on cadastral parcels in the urban area and urban cadastral maps mentioned this subject.

To clarify this subject an explanation was added to the new revision. “Since the juridical map and registered parcels are fixed properties and obtained from the special administrative procedures, these parcels cannot be changed and replaced by other parcels without legal processing. In this research, the base map from the Cadaster Department was produced by NCC and had no registered parcel. This base map is usually used beside another dataset for registered parcels to handle new requests for property registration. Hence, enriching this base map reduces the Cadastral Department process for registration. On the other hand, other organizations such as utilities and infrastructure organizations can effectively use the proposed framework to enrich their maps for better decision-making and urban management.”

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Thank you for submitting revision. My concerns about improving quality of the manuscript have been addressed. 

Minor editing is required

Reviewer 2 Report

The previous comments have been addressed.

Reviewer 3 Report

Let's say, you adjusted the purpose of the research. But remember, there is no changing of registered legal relationships "automatically".