Parrot Sequoia RGB Images Blurry?

Testing a Parrot Bluegrass with Sequoia Sensor.  Ran it at highest speed on a bright sunny day, 130 feet AGL.  The images for the other bands look like I would expect for that altitude and the sensor resolution. The detail is uniform throughout the image.  However the detail in the RGB images appear to be non-uniform with bands of blurriness.  (See example provided)  I have not idea if this is normal since I’ve never used this sensor before.  I would expect due to the tight integration between Pix4D and Sequoia that the software engineers wouldn’t provide a setting that would be essentially incompatible with the sensor (too fast of a flight speed).  But then that begs the question why allow there to be a setting at all?  Maybe the answer is, to slow it down if the images are blurry (LOL.)  

https://i.imgur.com/dP7bDIM.jpg

I need to do some actual shots where the lighting may be lower with overcast instead of sunny.  My gut is saying if the issue really is due to drone speed then it will be worse in lower lighting.  Can the ISO setting of the RGB sensor be raised or can all I do is lower the drone speed? Or is this how the images typically look and the fast setting is pretty reliable?

Was running at 75% overlap.  

Hi,

Sometimes, a noticeable rolling shutter effect can happen on Sequoia RGB camera. We usually recommend to fly higher and to reduce the speed of the UAV, in order to reduce the rolling shutter effect. If you are using a quadcopter, a gimbal would reduce the high frequency vibrations on the camera and might also attenuate this effect.

We use rolling shutter correction for the RGB sensor of sequoia, but if the speed is too much, the algorithm cannot correct for it very accurately. I would recommend to reduce the speed, that would give better images. If the images themselves are blurry the software will not produce very good outputs. I have got good results at 60 m flight height. For agriculture, we recommend 85% frontal overlap and 70% side overlap.

No way, I have already made flights at 120m and was not able to align the rgb photos on mapper for 3d models. Is there a specific workflow I am missing here. Note: I am talking about using the sequoia RGB to make 3D models, for 2D Ag workflow models the mapper can handle, but there is nothing to squeeze out of a 2d rgb model for agriculture

Gustavo, Are you not able to make a 3D model from sequoia RGB images? Can you explain a bit more about “was not able to align the rgb photos on mapper for 3d models.”? Few screenshots would be helpful

just check this report. probably one of teh best photo alignment I already achieved with the sequoia RGB camera for 3d processingchiapeta daninhas rgb_report.pdf (1.3 MB)

Did you try processing with the ag RGB template?

The RGB template is only for 2D products. On the description it says, “flat terrain and 2D maps”

Hi, For any RGB cameras when you fly on a homogeneous area, you should use ag RGB template. I assume your field is not on a hilly slope. The ag RGB template has been optimized so that you get better results for agricultural areas. If you need any extra outputs, you can always enable them. For example, use the ag RGB template and if you need 3D model, enable the generation of mesh:

You can do the same for point cloud, DSM, DTM etc.

So, probably I am not having good results due to the inclination of the terrain. It is not a flat terrain wich I am trying to map.

@sesekeh180 Can you write us a support ticket and share the dataset with us? We will try our best to help and diagnose why it does not work accurately.

It’s a very large and heavy dataset. Better stick to other programs for now, and keep trying different settings

Sometimes, a noticeable rolling shutter effect can happen on Sequoia RGB camera. If this is the issue you are facing with, we usually recommend to fly higher and to reduce the speed of the UAV, in order to reduce the rolling shutter effect. If you are using a quadcopter, a gimbal would reduce the high frequency vibrations on the camera and might also attenuate this effect.

The main use case of sequoai RGB is scouting crops and identifying how they grow. The focus of Sequoia mostly lies on delivering highly precise and accurate multispectral data. The RGB camera on Sequoia is an add-on, designed to generate overview images of agricultural fields.

What Sequoia`s RGB camera currently can be used for:

  • Creating overview images of agricultural fields in RGB that can support the multispectral data by offering a birds eye view
  • Creating orthomosaics that allow basic measurements of lengths and areas

What Sequoia`s RGB camera cannot be used for:

  • Creating point cloud and DSM data for 3D data analysis
  • Creating highly precise orthomosaics with the necessary precision that is needed e.g. for professional surveying

Hi Momtanu, I honestly believe that sequoia’s RBG can’t be used to anything at all.

We should not rely on any information generated by the orthos due to i’ts inconsistency. I would not recommend take measures from any product generated by the RGB camera today. Even when using the RGB on the fields software, you can see a lot of erros in the orthos with repeated objects dragged from one point to another.

Till we are not able to have a solid workflow, generating reliable results repeatedly, I would say that even for scouting this camera should not be used. The composite image generated by the multispectral sensors on the sequoia is far better to be used for anything, even having a 1,2Mb sensor.

1 Like

True because the multipsectral cameras in the sensor have global shutter but the RGB orthomosiac shouldnt look very bad as well. Maybe you can send us a datset sometime so taht we could try to solve this issue. I would also recommend flying higher and slower, that really helps.

Hi, Just my 2 cents on that subject…
I bought 3 sequoias for msp AND rgb mapping, mounted on Ebee’s with dedicated integration kit.

The documentation on that camera is misleading, as it present the RGB sensor as a quality one, with a better GSD that multispectral ones.

All my RGB surveys are unusable, the images have this oil painting effect and even with 85/90 overlap they can’t calibrate correctly. Nice results with msp picture however.

Parrot should explain it clearly before selling that 4k € cam…

Most likely, your camera does not have enough pixels to transmit all colors over long distances.

Please contact a PM.

Merci pour les précisions. Mais j’aimerais savoir si vous avez choisi la vitesse NORMALE OU LENTE ?? . Moi j’ai un bluegrass et j’ai aussi les même problèmes avec le capteur RVB.

Parcontre même avec un chevauchements de 60% en frontal comme en latéral à 30m ou 50m j’ai de bon résultat avec le capteur multi spectrale.

Multispectral bands (like green and NIR) have more variability then RGB images for fields/agriculture. You will need more overlap. The recommended overlap is 85% frontal and 70% overlap. You will need to fly slow and higher to reduce the distortion from rolling shutter

Great. I Will try it.