Sequioa multispectral reflectance: greater than 1?

PIX4Dmapper PIX4Dmapper Questions/Troubleshooting
1

I had thought that reflectance vaues could only be from 0-1, since it is the ratio of the incident to reflected radiation. However, I have a project where the red edge reflectance varies from 0-2.8.

The other 3 channels give realistic results - between 0 and 0.6. With ‘no correction,’ selected in the Radiometric Processing and Calibration section of the Index Calculator the raw values for each sensor all have similar mean, min. and max. values. With all other corrections the red edge values are significantly higher (3x) than the others and greater than one.

On subsequent flights, all four sensors have given values in the expected range.

Does anyone have any suggestion about why this might be the case and what might be the best way to fix the problem?

2

Hey Iain,

Green, Red, Red Edge and NIR should normally be between 0 and 1 and NDVI could be between -1 and 1. Sequoia takes three sets of calibration images so I would switch the calibration images. I would also make sure that there are no scratches or the calibration target is not damaged where the auto-detected boundary is. If that is the case I would make the calibration target boundary smaller manually. If that doesn’t work I would use the Camera Only option. 

3

Thanks Selim. I’ve now tried that but there still seems to be issues, even with the camera only setting. The actual results (in term of spatial variation) look good, its just the scaling of the reflectance that is off. Thankfully from my perspective its the spatial variation that is important for this flight and not the actual values.

4

Hey Iain,

If the variation looks okay and you are not doing a comparison over time you may be okay but if you are looking to compare the data over time the data that you have may not be reliable.

5

Thankfully in this instance temporal variartion isnt necessary. Can you think of any other reason why i may not be able to get it calibrated properly? I’ve not had this issue on subsequent flights so it’s not a sensor issue and I’ve followed the emotion/pix4d process.

6

Hey Iain,

Is it possible that there is an outlier in the field that is causing inaccurate reflectance values? Also the calibration target is not damaged or the images are not under or over exposed? I would look up the reflectance values in a histogram either in QGIS or ArcMap. You would be able to determine the frequency of those inaccurate values, if it is very little, it probably is an outlier if it is frequent it may be more concerning. You could also try to load only the RedEdge images and process it by itself, to see if it makes a difference. Moreover you could export reflectance values as point . shp in Pix4D if you want to further investigate certain regions. 

7

Hi Iain,

If using a calibration target one should always expect reflectance values between 0 and 1. However, it may be that the sensor has defects or the connection is corrupted hence it is expected in this situation to have wrong values.To check this you should use the Camera Only correction and the calibration target. Then if the values are okay that means that there may be something wrong with the data from the sunshine. If not then please send us the dataset. 

Could you do this test and let me know how it went? To better understand you could post a screenshot of the index calculator with the reflectance map of the red-edge and the statistics on the left sidebar.

 

Thanks,

Ina

 

8

Thanks Ina,

I used a calibration target. The sensor shouldn’t have any defects as in three subsequent flights the data has all been in the expected range.

Firstly here is a map of the uncorrected values:

Secondly, here is a screenshot of the case with the camera only correction:

You can see that the other three sensors are giving reasonable results. When I include sun irradience the max. reflectance drops slighty (to about 2.8).

Is it best if I send you the dataset? If so what is the best way to send it? I am out of the office tomorrow so will send on Friday.

Thanks very much,

Iain

9

 Hi Iain,

 

Yes please send it to me so I can take a look. The best way for sharing would be our cloud platform.You can find the instructions here. 

 

Cheers,

Ina