Excellent video, with *one minor* adjustment: Metashape's re-alignment regression does not use the accuracy settings as static limits. Rather, they are only used as *weights relative to one another.* For example, setting photos to 20m and GCP to 1cm is a 2000:1 weight for GCP's. Setting photos to 10m and GCP to 5mm (defaults) is *an identical 2000:1 weight.* When doing the regression (similar to least squares) to determine the most likely coordinate, points are evaluated based on their weight. The static number itself means nothing, only the relative difference is consequential. To explain more simply, imagine three UAV's were used to survey. One is corrected via PPK, one with RTK, and one traditional GNSS w/o correction. In that instance, you could set GNSS to 100m, RTK to 10m, PPK to 8m, and GCP to 10cm. The isolated numbers are not consequential; however, the *relative* weight will be 10,000 times higher for GCP's than GNSS, while RTK is 10 times higher weight than uncorrected GNSS, and PPK is about 25% higher weight than RTK. This helps you understand that with a single image set and GCP's, only one number needs to be adjusted and that the effect of doing so is logarithmic, for example, changing 1cm to 3cm is proportionately equal with 3cm to 9cm.
Hi, thanks for video. What I see is that you do not take the full advantages of a RTK drone into account. Instead you weight and still rely on multiple GCPs. I think it's better to import the accuracy values for each image (see settings import accuracy) and process orientation based on this RTK image coordinates. After this, import and mark 3+ GCPs, uncheck image coordinates and use the "update" tool to transform to to GCPs. This will avoid unwanted bending of the model. Cheers, Hans
Hi Hans, thanks for the comment. This exclusion was done intentionally. I find that the majority of viewers do not have or do not use RTK enabled drones and thus this was to clearly illustrate what Metashape does with control points if left unchecked and unmanaged. You are of course correct that when we use RTK (more specifically PP-RTK) positions, these should be taken into consideration and used to guide the orientation. RTK is only accurate in a relative sense unless setup on a known points and observed for extended time periods, hence the need to use the "Update" function as you have pointed out. If it was possible, I'd only ever use PP-RTK and then even the "Update" function should become unnecessary. Keep in mind too that ALL positions are only as good as the equipment being used and blind faith in any one of the technologies being used is a risk. If you and the team at RedCatch would like to supply some data and we can make a video together to show the best way to harness RTK, specifically with RedCatch, please do get in touch! Thanks
A minor point at 1:34 you talk about microns, but that calculation is actually in pixels. That is, the units cancel out when dividing. This is so that the everything the CV pipeline deals with is in pixels which makes the maths much easier.
I'm always excited when I see a new video on this channel. This one is especially good because there are so few and perhaps no good videos on this topic. I can't watch this right now but I'm excited to get to it tonight. Thanks! The Metashape instructional materials seem pretty poor to me. They should hire you to produce good work on their channel... (if you would be willing and able to do that)
Hi! Sorry I've been pretty quiet this year. It's been a bit of a challenging year so far, but things are looking up and I plan to add to the channel soon and hope it's of some value to you all! Thanks for checking in :)
Mark, can you tell me if you use a desktop or laptop for most of your map processing? What you recommend. Your channel is the best I’ve come across and I’ve been through it all. Much appreciation to you and your work.
Hey TexasHunter Thanks for the kind comments. I use a desktop for the processing work, even though mine is pretty old now, it’s reliable in a way I don’t think laptops are if pushed hard over long periods. I think a good GPU with 8GB worth of memory and at least 64GB of RAM are where you want to be for most jobs. You can use lesser specs, but you may find processes start to fail. (More RAM would be great, 128GB and up) Your CPU will dictate the speed with which certain tasks run, but I can’t recall seeing smaller CPUs causing failures like small GPUs and low RAM would. Hope that helps a little.
Great lesson, as usual. @inspiredScience"s comment is pertinent as well. Also, in my use, I usually get slightly reduced camera errors (rtk drone images) when I also select "fit additional corrections". I think this is just letting MS tweak the camera calibration based on the additional data. Have you tried using it in your testing? thanks again! From the Manual "Fit additional corrections: With this option enabled Metashape estimates additional coefficients that are necessary to achieve better accuracy."
Hey David - thanks for the comments. You are very correct, @InspireSciences comments were exactly right, I've just been away for a while, but back now and will answer that comment too. I haven't tried the "fit additional corrections" options too many times, to be honest, and only approach it when things don't seem to be going well. But it is good advice and well worth a look. Thanks!
Would you consider a focal length change of 4404 initial to 4235 adjusted to be too large of a change? This is just after aligning the photos on High accuracy. It drops to 4222 after doing all the tie point filtering and optimization. I have cross referenced the accuracy of the RTK drone surveys to a ground based LIDAR scanner and found the drone surveys to be very accurate compared to that, so not sure if I should be worried about the focal length change.
Great video and good explanation! When accuracy values are included with the GCP file, eg they are provided in the export from the GNSS rover, and accuracy columns are selected during import into metashape, I guess those accuracy values override the reference settings for marker accuracy? Would the same apply to camera accuracy when rtk is being used and values are loaded from xmp data? Tools > Preferences > Advanced > Load Camera location accuracy from XMP meta data
Thanks for watching. My opinion (may be a little off) is this: I'd be skeptical to leave the marker accuracies as they are from the receiver, unless they have been post-processed. The hassle is that the real-time corrections are based on expected inputs, and don't always include drift and other changes in the GPS satellites, or something else that may have changed at the CORS station - I'd say use the accuracy values with caution, but allow them a larger error potential and see what the result is. You will find that the error stops increasing at a certain point if your points are good. The drone accuracies, if you use RTK, would be valid in a relative sense. Like in the video example, we began with 18cm but ended up at 1m (even though I allowed 3m) because RTK is not absolutely positioned, but it is only relative to the RTK base. (If you setup the RTK base on a known point and enter that value into the RTK receiver, different story) In that case the image-to-image alignment was an average of 18cm, but 1m from known points on the ground. Metashape doesn't allow you to easily see both values at the same time. A good idea is to use those values as a starting point and take it from there, paying close attention to changes as you process. As you use your own system more, I expect you'll get a feeling for what is correct for you. If you can use PPK however, that's a different story and I'd likely trust those values a lot more.
Dear, congratulations on the video... I have a question and maybe you can help me: In the Ground Control Points section of the Metashape accuracy report, I have the following Error X (cm) 2.33 Y error (cm) 2.07 Z error (cm) 1.15 XY (cm) 3.12 Total (cm) 3.33 My orthomosaic has a pixel measuring 1.72 cm and the DEM pixel measuring 3.44 cm If I measure the length of an object in orthomosaic, my error will be between (1x GSD) and (1.5 x GSD), right? In this calculation, should I consider the pixel of 1.72 cm in the orthomosaic or 3.12 cm in XY of the report? Is XY error only for area measurement? If I measure the height of the object, I should calculate the error between 2 to 3xGSD, right? Should I consider the 3.44 cm pixel in the DEM or 1.15 cm in Z indicated in the report? Thanks.
Regarding your tie-point cleaning process, the other guide you linked uses different settings for the 'Optimize Camera Alignment' prompt. Can you clarify which video has the boxes checked?
Hi - If you have images captured at various altitudes and/or you have oblique images along with those captured at nadir, then I would use the "Adaptive Model" option when optimizing the camera. However, if you have a more simple dataset where images are captured at a single altitude and generally all at nadir, then you do not use adaptive modelling and you would have the boxes checked per the default setting. I hope that helps.
I would just use the optimization. I think that if you re-run the alignment you "force" the images onto the control position and therefore deviate away from a true AT result. Thus, you can't report it as an accuracy value using those points because you're fixing them. If however you have many additional check points that you only add afterwards, then it should be okay.
Guten Tag, da ich gesehen habe, dass Sie die schweizer Landeskoordinaten gebraucht haben, gehe ich davon aus, dass Sie auch aus der Schweiz sind? Ich hätte eine frage betreffend LV95 Geoid und dem Import von dxf…. Würde mich rieseig über eine kurzen Austausch freuen! Freundliche Grüsse
Hello! When I calc the focal length in micron, I am not getting the initial calculated number in Metashape. For example: 12.29 mm length divided by .00335821 = 3,659.68. The initial calculated length in micron is 3713.29. Im not sure where the discrepancy lies.
Hey Dennis! Question: The 3713 value form Metashape, is that before or after aligning images? I expect it'll change a little after your first alignment step and this difference isn't outrageous. But, if this is the value that Metashape gives right after adding your data, then I am not sure what is happening as your maths is correct.
@@geospatialtips This is before I aligned the images. After aligning, the adjusted value is F= 3705.749. So, what I did was start a new project, upload images and then manually input F= 3713.29 as the initial. Not sure if this is correct but after aligning so far everything seems to be in an acceptable range. I just saw that you do one on one tutoring type sessions. Im interested in setting something up soon!
@@geospatialtips Fumbling my words here trying to explain. Ill recap here quickly incase anyone else is having the same issue. After importing images I went into the Camera Calibration menu and noticed that it had a value already populated (3713.29). Then I calc'ed myself using the equation 12.29 (Mavic 3 Enterprise wide lens focal length) X pixel size .00335821 = 3,659.68. After attempting to align the images, I went back into the Camera Calibration menu and noticed that the value did not change to the correct length in micron (3,659.68). So I opened a new project and started over, this time after importing the images I went back to the menu and manually entered the calc'ed figure. Then I aligned the images and my adjusted value is F= 3705.74, and this is within the acceptable range! Im not sure why Metashape is giving the wrong calculation after adding the images.
hi i need an accurate Ortho mosaic of all data for cad work so is it necessory to follow gcp method or i can make without gcp also an accurate orthomosaic image as you know we can align in autocad easily without gcp ortho
Hi - it will depend on how accurate you need the end result to be. Also, is the drone you are using RTK enabled, or are you using just the onboard GPS? If only the onboard GPS it will be less accurate. Using this ortho is still possible, but you have no way of ensuring accuracy, so we get back to the question of, "How accurate do you need it to be?"
Hi - check the Metashape Preferences/Advanced tab and make sure the box is ticked to allow Metashape to read that kind of data on import. If it still is not there, right click on one of your images in File Explorer, go to Properties and then Details. Scroll down the list and make sure the coordinates are saved in the raw image data. If none of that helps you, maybe send me one of your images and I will take a look. (geospatiltips@gmail.com)
Hello, I would like to count plants in a field based on their height. I think that this should be possible using a dense cloud or a 3D model. What do you think, can this be done or would another way be better?
Yes, I think this is possible. First you would need to establish the height of the plants and you can do this in two ways I can think of. 1. Calculate the height of the points (of the plant) or that section of the model above the surrounding ground points and do this for each plant. Add an attribute for each plant and perform the count once all plants are measured. This method would be time consuming. 2. Classify the ground points then reduce all ground points to 0m elevation. Allow all data to shift with it. The result will be that all ground points are 0m and all other points are now their own height above the ground i.e. the height of each plant. Therefore you could now only use data within your selected height range and delete of hide all other data. If you wanted only plants taller than 2m, you delete everything below 2m. Or a plant taller than 1m but shorted than 1.5m, delete all data below 1m and all data above 1.5m. The result is you will have points of data that should be the plants you need and then you can count of of these points of data. To make it easier, groups the points or pixels by proximity so you only have 1 group per plant, then select all plants. It should work.
Hi, Iwould like to know. If I can do the process of aligning the coordinate system before any process mentioned in the video. Can I work with the correct coordinate system from the beginning? Or do I need to do this process only after aligning the photos?
Guten Tag, da ich gesehen habe, dass Sie die schweizer Landeskoordinaten gebraucht haben, gehe ich davon aus, dass Sie auch aus der Schweiz sind? Ich hätte eine frage betreffend LV95 Geoid und dem Import von dxf…. Würde mich rieseig über eine kurzen Austausch freuen! Freundliche Grüsse
Hi - no, I'm not Swiss, I am South African. I just happened to use a Swiss dataset for this video. I'm happy to help you if I can, you can email me at geospatialtips@gmail.com if you like.
Absolute goldmine of a channel!
As someone who deals with Metashape often, your videos are super helpful.
Excellent video, with *one minor* adjustment: Metashape's re-alignment regression does not use the accuracy settings as static limits. Rather, they are only used as *weights relative to one another.* For example, setting photos to 20m and GCP to 1cm is a 2000:1 weight for GCP's. Setting photos to 10m and GCP to 5mm (defaults) is *an identical 2000:1 weight.*
When doing the regression (similar to least squares) to determine the most likely coordinate, points are evaluated based on their weight. The static number itself means nothing, only the relative difference is consequential.
To explain more simply, imagine three UAV's were used to survey. One is corrected via PPK, one with RTK, and one traditional GNSS w/o correction. In that instance, you could set GNSS to 100m, RTK to 10m, PPK to 8m, and GCP to 10cm. The isolated numbers are not consequential; however, the *relative* weight will be 10,000 times higher for GCP's than GNSS, while RTK is 10 times higher weight than uncorrected GNSS, and PPK is about 25% higher weight than RTK.
This helps you understand that with a single image set and GCP's, only one number needs to be adjusted and that the effect of doing so is logarithmic, for example, changing 1cm to 3cm is proportionately equal with 3cm to 9cm.
Hi, thanks for video. What I see is that you do not take the full advantages of a RTK drone into account. Instead you weight and still rely on multiple GCPs. I think it's better to import the accuracy values for each image (see settings import accuracy) and process orientation based on this RTK image coordinates. After this, import and mark 3+ GCPs, uncheck image coordinates and use the "update" tool to transform to to GCPs. This will avoid unwanted bending of the model. Cheers, Hans
Hi Hans, thanks for the comment. This exclusion was done intentionally. I find that the majority of viewers do not have or do not use RTK enabled drones and thus this was to clearly illustrate what Metashape does with control points if left unchecked and unmanaged.
You are of course correct that when we use RTK (more specifically PP-RTK) positions, these should be taken into consideration and used to guide the orientation. RTK is only accurate in a relative sense unless setup on a known points and observed for extended time periods, hence the need to use the "Update" function as you have pointed out. If it was possible, I'd only ever use PP-RTK and then even the "Update" function should become unnecessary.
Keep in mind too that ALL positions are only as good as the equipment being used and blind faith in any one of the technologies being used is a risk.
If you and the team at RedCatch would like to supply some data and we can make a video together to show the best way to harness RTK, specifically with RedCatch, please do get in touch!
Thanks
A minor point at 1:34 you talk about microns, but that calculation is actually in pixels. That is, the units cancel out when dividing. This is so that the everything the CV pipeline deals with is in pixels which makes the maths much easier.
I'm always excited when I see a new video on this channel. This one is especially good because there are so few and perhaps no good videos on this topic. I can't watch this right now but I'm excited to get to it tonight. Thanks!
The Metashape instructional materials seem pretty poor to me.
They should hire you to produce good work on their channel... (if you would be willing and able to do that)
If I wasn't an idio*! I would have found 99 videos on this subject in other languages!
great video, thank you. can you explain a bit, what was the process and equipment used for input data?
Very nice. Congrats my friend. Liked. All the best, from Brazil
I keep checking to see if there's any new videos.
Hope everything is good at the geospatial tips Home Office 🙂
Hi! Sorry I've been pretty quiet this year. It's been a bit of a challenging year so far, but things are looking up and I plan to add to the channel soon and hope it's of some value to you all! Thanks for checking in :)
@@geospatialtips
Sorry to hear about the challenges and I hope everybody is okay.
Glad to hear that things are looking up!
Mark, can you tell me if you use a desktop or laptop for most of your map processing? What you recommend. Your channel is the best I’ve come across and I’ve been through it all. Much appreciation to you and your work.
Hey TexasHunter
Thanks for the kind comments.
I use a desktop for the processing work, even though mine is pretty old now, it’s reliable in a way I don’t think laptops are if pushed hard over long periods.
I think a good GPU with 8GB worth of memory and at least 64GB of RAM are where you want to be for most jobs. You can use lesser specs, but you may find processes start to fail. (More RAM would be great, 128GB and up) Your CPU will dictate the speed with which certain tasks run, but I can’t recall seeing smaller CPUs causing failures like small GPUs and low RAM would.
Hope that helps a little.
@@geospatialtips that’s exactly the advice I was looking for. Thank you sir
Great lesson, as usual. @inspiredScience"s comment is pertinent as well. Also, in my use, I usually get slightly reduced camera errors (rtk drone images) when I also select "fit additional corrections". I think this is just letting MS tweak the camera calibration based on the additional data. Have you tried using it in your testing? thanks again!
From the Manual "Fit additional corrections:
With this option enabled Metashape estimates additional coefficients that are necessary to achieve
better accuracy."
Hey David - thanks for the comments. You are very correct, @InspireSciences comments were exactly right, I've just been away for a while, but back now and will answer that comment too.
I haven't tried the "fit additional corrections" options too many times, to be honest, and only approach it when things don't seem to be going well. But it is good advice and well worth a look. Thanks!
Would you consider a focal length change of 4404 initial to 4235 adjusted to be too large of a change? This is just after aligning the photos on High accuracy. It drops to 4222 after doing all the tie point filtering and optimization. I have cross referenced the accuracy of the RTK drone surveys to a ground based LIDAR scanner and found the drone surveys to be very accurate compared to that, so not sure if I should be worried about the focal length change.
Great video and good explanation! When accuracy values are included with the GCP file, eg they are provided in the export from the GNSS rover, and accuracy columns are selected during import into metashape, I guess those accuracy values override the reference settings for marker accuracy? Would the same apply to camera accuracy when rtk is being used and values are loaded from xmp data?
Tools > Preferences > Advanced > Load Camera location accuracy from XMP meta data
Thanks for watching.
My opinion (may be a little off) is this:
I'd be skeptical to leave the marker accuracies as they are from the receiver, unless they have been post-processed. The hassle is that the real-time corrections are based on expected inputs, and don't always include drift and other changes in the GPS satellites, or something else that may have changed at the CORS station - I'd say use the accuracy values with caution, but allow them a larger error potential and see what the result is. You will find that the error stops increasing at a certain point if your points are good.
The drone accuracies, if you use RTK, would be valid in a relative sense. Like in the video example, we began with 18cm but ended up at 1m (even though I allowed 3m) because RTK is not absolutely positioned, but it is only relative to the RTK base. (If you setup the RTK base on a known point and enter that value into the RTK receiver, different story) In that case the image-to-image alignment was an average of 18cm, but 1m from known points on the ground. Metashape doesn't allow you to easily see both values at the same time.
A good idea is to use those values as a starting point and take it from there, paying close attention to changes as you process. As you use your own system more, I expect you'll get a feeling for what is correct for you.
If you can use PPK however, that's a different story and I'd likely trust those values a lot more.
Dear, congratulations on the video...
I have a question and maybe you can help me:
In the Ground Control Points section of the Metashape accuracy report, I have the following
Error X (cm) 2.33
Y error (cm) 2.07
Z error (cm) 1.15
XY (cm) 3.12
Total (cm) 3.33
My orthomosaic has a pixel measuring 1.72 cm and the DEM pixel measuring 3.44 cm
If I measure the length of an object in orthomosaic, my error will be between (1x GSD) and (1.5 x GSD), right?
In this calculation, should I consider the pixel of 1.72 cm in the orthomosaic or 3.12 cm in XY of the report?
Is XY error only for area measurement?
If I measure the height of the object, I should calculate the error between 2 to 3xGSD, right?
Should I consider the 3.44 cm pixel in the DEM or 1.15 cm in Z indicated in the report?
Thanks.
Regarding your tie-point cleaning process, the other guide you linked uses different settings for the 'Optimize Camera Alignment' prompt. Can you clarify which video has the boxes checked?
Hi - If you have images captured at various altitudes and/or you have oblique images along with those captured at nadir, then I would use the "Adaptive Model" option when optimizing the camera. However, if you have a more simple dataset where images are captured at a single altitude and generally all at nadir, then you do not use adaptive modelling and you would have the boxes checked per the default setting. I hope that helps.
@@geospatialtips It helps a lot! Thanks for the replies 👍
thnks
What does it mean when I have more error in my control points compared to my checks? (Talking about 5mm difference)
Great video! After putting GCP I must re run alignment or using optimazation is enough?
I would just use the optimization. I think that if you re-run the alignment you "force" the images onto the control position and therefore deviate away from a true AT result. Thus, you can't report it as an accuracy value using those points because you're fixing them. If however you have many additional check points that you only add afterwards, then it should be okay.
Guten Tag, da ich gesehen habe, dass Sie die schweizer Landeskoordinaten gebraucht haben, gehe ich davon aus, dass Sie auch aus der Schweiz sind?
Ich hätte eine frage betreffend LV95 Geoid und dem Import von dxf….
Würde mich rieseig über eine kurzen Austausch freuen!
Freundliche Grüsse
Nice video❤
Hello! When I calc the focal length in micron, I am not getting the initial calculated number in Metashape. For example: 12.29 mm length divided by .00335821 = 3,659.68. The initial calculated length in micron is 3713.29. Im not sure where the discrepancy lies.
Hey Dennis!
Question: The 3713 value form Metashape, is that before or after aligning images? I expect it'll change a little after your first alignment step and this difference isn't outrageous. But, if this is the value that Metashape gives right after adding your data, then I am not sure what is happening as your maths is correct.
@@geospatialtips This is before I aligned the images. After aligning, the adjusted value is F= 3705.749. So, what I did was start a new project, upload images and then manually input F= 3713.29 as the initial. Not sure if this is correct but after aligning so far everything seems to be in an acceptable range. I just saw that you do one on one tutoring type sessions. Im interested in setting something up soon!
@@geospatialtips correction; the value I manually entered prior to aligning was the 3659.69.
@@geospatialtips Fumbling my words here trying to explain. Ill recap here quickly incase anyone else is having the same issue. After importing images I went into the Camera Calibration menu and noticed that it had a value already populated (3713.29). Then I calc'ed myself using the equation 12.29 (Mavic 3 Enterprise wide lens focal length) X pixel size .00335821 = 3,659.68. After attempting to align the images, I went back into the Camera Calibration menu and noticed that the value did not change to the correct length in micron (3,659.68). So I opened a new project and started over, this time after importing the images I went back to the menu and manually entered the calc'ed figure. Then I aligned the images and my adjusted value is F= 3705.74, and this is within the acceptable range! Im not sure why Metashape is giving the wrong calculation after adding the images.
hi i need an accurate Ortho mosaic of all data for cad work so is it necessory to follow gcp method or i can make without gcp also an accurate orthomosaic image as you know we can align in autocad easily without gcp ortho
Hi - it will depend on how accurate you need the end result to be. Also, is the drone you are using RTK enabled, or are you using just the onboard GPS? If only the onboard GPS it will be less accurate. Using this ortho is still possible, but you have no way of ensuring accuracy, so we get back to the question of, "How accurate do you need it to be?"
Im having trouble. My cameras do not have any coordinates tied to the photo. How is the process different. My elevation is way off.
Hi - check the Metashape Preferences/Advanced tab and make sure the box is ticked to allow Metashape to read that kind of data on import.
If it still is not there, right click on one of your images in File Explorer, go to Properties and then Details. Scroll down the list and make sure the coordinates are saved in the raw image data.
If none of that helps you, maybe send me one of your images and I will take a look. (geospatiltips@gmail.com)
Hello, I would like to count plants in a field based on their height. I think that this should be possible using a dense cloud or a 3D model. What do you think, can this be done or would another way be better?
Yes, I think this is possible.
First you would need to establish the height of the plants and you can do this in two ways I can think of.
1. Calculate the height of the points (of the plant) or that section of the model above the surrounding ground points and do this for each plant. Add an attribute for each plant and perform the count once all plants are measured. This method would be time consuming.
2. Classify the ground points then reduce all ground points to 0m elevation. Allow all data to shift with it. The result will be that all ground points are 0m and all other points are now their own height above the ground i.e. the height of each plant. Therefore you could now only use data within your selected height range and delete of hide all other data. If you wanted only plants taller than 2m, you delete everything below 2m. Or a plant taller than 1m but shorted than 1.5m, delete all data below 1m and all data above 1.5m. The result is you will have points of data that should be the plants you need and then you can count of of these points of data. To make it easier, groups the points or pixels by proximity so you only have 1 group per plant, then select all plants. It should work.
Hi, Iwould like to know. If I can do the process of aligning the coordinate system before any process mentioned in the video. Can I work with the correct coordinate system from the beginning? Or do I need to do this process only after aligning the photos?
Yes, you can set it beforehand as well and use your desired system from the beginning. The process would be much the same.
Guten Tag, da ich gesehen habe, dass Sie die schweizer Landeskoordinaten gebraucht haben, gehe ich davon aus, dass Sie auch aus der Schweiz sind?
Ich hätte eine frage betreffend LV95 Geoid und dem Import von dxf….
Würde mich rieseig über eine kurzen Austausch freuen!
Freundliche Grüsse
Hi - no, I'm not Swiss, I am South African. I just happened to use a Swiss dataset for this video.
I'm happy to help you if I can, you can email me at geospatialtips@gmail.com if you like.