Crowd Density Estimation and Mapping Method Based on Surveillance Video and GIS

Round 1

Reviewer 1 Report

This study combined GIS and video surveillance, and used intelligent video analysis technology to achieve intelligent detection and mapping of crowd density in video scenes under the geospatial reference. The proposed method has great application value in the field of public security. The manuscript was well organized and the results are fully demonstrated. However, the following minor problems need to be solved. It is suggested to accept this manuscript after addressing these problems.

1.     The method part is too detailed, and it is recommended to simplify it appropriately.

2.     Dis, Inc, etc. in Figure 1 should be declared with full name in the figure or figure caption.

3.     There are too many figures. It is recommended to combine Figure 1 - Figure 8 into 2-3 figures.

4.     The discussion needs to be strengthened.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The proposed algorithm for estimating the number of people from an area versus identifying heads from an image is functional and successful. However, it is not entirely clear to me

A) to what extent the camera angle and distance, from which the main variable - area - is derived, are involved in the classification. Clearly these are dependent variables. 

B) I understand that the authors want to remove the influence of perspective but I am not clear about the role of GIS in this process, nor am I clear about the reason for using homography when I can determine the real coordinates anyway thanks to topographic reference and apply a polynomial transformation to identical points. The image rectification is based on identical points so the area could be derived directly from the transformation of the camera image to topographic data (derived from a map or image).

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The paper is very interesting, and it focused the crowd counting with vedio and GIS image. The paper provided a framework of crowd counting and implemented with case tests. 

just some comments and further discussion. 

1.  What kind image did the paper use? The flowchart in Figure 1 mentioned high-definition remote sensing image. Orthoimage or  oblique image? How to match the vedio with it?  Alo I noticed the statement in Section 3.3.2. 

2. Is the parameter "inclination" a key factor that influence the resut of calculation and prediction? Different inclination would cause different overlay of the individuals in the crowd and may effect the  estimation of the count. Did the author test the different inclinations of the vedios?

3.  All tests are in the same place. Different places with test can illustrate the method's robustness. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx