Support Website Contact Support Blog

Cloud shadows on orthomosaic using multispectral camera

I a having an issue with how Pix4d selects images to use for the orthomosaque where there is cloud shadow in some of the images. I am using an  “Airinov multispec4C” camera which Pix4D recognizes as known camera model. This camera can recorded four bands (Green,Red,Red-edge and NIR) and in good conditions has provided good results.

The issue we encountered is that although Pix4dD is stitching each band successfully, when there is cloud shadow in some of the images the location of the clouds between the stitched bands is different. The can have a dramatic impact on an NDVI. I have attached screen shots of the ‘Red’, ‘NIR’ and ‘NDVI’ from a test site where this was observed.

Is there a way in Pix4D to ensure that when the band for one image is used for the orthomosaque, the other bands from the same image will also be used in the same location in their corresponding orthomosaque? I also tried to submit this as a request but the site kept crashing.

Thank you for any help you can provide.

 

 

NDVI

 

NIR

 

Red

Dear Timothy,

It is not recommended to obtain imagery in scattered clouds conditions. This will introduce artefacts that might be hard to detect (though in your particular case they are easier to see). Indeed the reflected light will depend on whether the area photographed is in the shade or in the sun. If one has a irradiance sensor to adjust the images, one could also imagine a scenario where the sensor is in the shade, but it takes an image of a sunny region. This will create problems with the radiometric corrections.

We strongly recommend obtaining the imagery either in clear sky conditions, or with a uniform cloud cover to obtain accurate data.

The value of a given pixel in the reflectance map is a weighted average of all the pixel values projecting on this reflectance map pixel. As such, all images are used, though the coefficient of a given image in the weighted average might vary from band to band.

Best regards,

Hello I would like to know how can i get to see the NDVI value of a given pixel once the index map is generated.

Also it is my understanding that I can hope to get a 1-3 pixel accuracy of a project, is that true also for an index map? As I am interested in overlapping layers with a given pixel size (1 m x 1 m) I would need to know that the NDVI value in that pixel is accurate. I am not sure what would be better if building an index map with a 1 m GSD or use a smaller value of resolution. 

Thanks.

Hi CmLab, 

In the bottom right corner of the Index Calculator view the value of the pixel that is currently under the mouse pointer is displayed. As shown in this screenshot: 


Otherwise, a great way to visualize the Index Maps and to know the NDVI values of a given area is to define a specific color for a certain range of value. More ressources on how the Index Calculator can be used are in this article from our Knowledge Base: https://support.pix4d.com/hc/en-us/articles/202558729 

The relative accuracy of a project reconstruction is between 1 to 3 times the Ground Sampling Distance (GSD) of the project. Note that these are expected values, but that the values in the project can be smaller or higher depending on the parameters of the project. This also applies to the Index Map. More about accuracy in this article: https://support.pix4d.com/hc/en-us/articles/202558889

The accuracy of the reconstruction depends on the GSD of the project. Hence, it will be the same ratio (1-3x GSD) if the GSD is at 10cm or at 1m, except that the scale changes. 

If you know that you need a resolution of the Index Map at 1m for your application, but you have the possibility to acquire data that has a GSD at a lower value (e.g. 10cm). We would suggest to acquire the data at a lower GSD and to downsample the Index Map to the desired resolution afterwards. This could reduce the impact of the expected reconstruction accuracy on the final value at a lower resolution. The resolution of the Index Maps can be defined in the Processing Options:

Regards