Quick Camera test: DJI Phantom 4 pro vs Mavic pro

Here are some quick results from the drone camera tests we did two days ago. We flow four different cameras under similar conditions. The data set should ideally be used buy a student to do some project and a deeper evaluation of the cameras. But as we know that many in the forest industri are thinking about the difference between the Dji Phantom 4pro camera with a global shutter and a larger sensor compared to the Dji Mavic pro which is smaller, cheeper and has a, for photogrammetry, poorer camera.

The drones (and cameras) was flown in the evening with low sun behind thin clouds, generating no shadows but also not bright light. This lighting condition should push the cameras a bit and at the same time give less problem with bright sunlit tree crowns against a dark shadowed ground.

A orthophoto was made in Pix4D cloud service using default settings. Here are the results:

The Phantom 4 pro camera (on the left) has a higher dynamic range in the spectral resolution and is also less over exposed on the road compared with the Mavic pro (right image).

For the final test the Sony a5100 and the Parrot sequoia needs to be added and the images needs to be compared in more depth as well as their performance when generating point clouds.

 

Three drone systems, four different cameras

We decided to take our drones to the forest and test some cameras today. We flow two 3DR Solo; one with Parrot Sequoia and one with a new Escadrone SoloMapper (Sony QX1 camera) and two DJI drones; one Phantom 4 pro and one Mavic pro. This was done because we had a great opportunity to acquire data from four forest patches that has just been scanned with a good terrestrial laser scanner. The scans was multi-scans with 16 scans in a 8 x 2 grid. The point cloud from the laser could be used for evaluating the point clouds generated from the drone cameras.

We tried to fly slow (5 m/s) and with lots of overlap (90/90). Sadly we ran in to trouble with the new SoloMapper, it did not acquire more then 15 images and then the system stopped. This was the first time ever we tested the camera system on the drone so it was probably a installation error of ours. We switch camera to a Sony a5100 that we have used before but which doesn’t have a gimbal, so we got good quality images.

This data set, with four different cameras, would be perfect for some student to use for a project or thesis.

 

SLUs drones goes thermal!

3DR Solo, equipped with Flir Vue and GoPro over prescribed fire.

The scientists at the Ljungberg Lab at the Swedish University of agricultural Sciences in Umeå went for a two day field excursion to test thermal cameras and new drones. Joining the students at the Fire Management course at the Forest faculty, who were going to make a prescribed burning of a 20 ha clear-cut. We wanted to test our new Solo helicopter drone from 3Drobotics and to capture thermal video and images from our Flir Vue camera. We also wanted to test the thermal camera in our fixed wing Smartplane drone.

We started by capturing RGB images from 200 meters above ground with the Smartplane before the fire was started. From this imagery we created a 7 cm Ground Sampling Distance (GSD) orthorectified image mosaic. Which could be used for describing the pre-fire state of the area.

Pre-fire orthophoto original wirh 7 cm resolution made from Smartplanes mapping system
Pre-fire orthophoto original wirh 7 cm resolution made from Smartplanes mapping system

We started by capturing video with both visual (GoPro, with modified lens) and thermal video (Flir Vue Pro 13mm 640×480) of the controlled burning (prescribed) of the clear-cut.

DSC04937
3DR Solo with GoPro and Flir Vue Pro, the cameras are tilted forward to acquire oblique video.

The video streams were later synchronized and fused to a side-by-side video using a software made by engineering students (an earlier project at the Ljungberg Lab).


Forest fire, fusion of thermal and RGB camera


Forest fire, thermal and RGB camera side-by-side

Under the duration of the controlled burning we flew the thermal camera multiple times, with the purpose of acquiring aerial thermal and visual (RGB) images to describe the burn process and to test the usefulness of having a thermal camera to find hotspots or ground fire hours after the fire front have passed an area.

3DR Solo ready to photograph controlled burning with thermal (Flir Vue) and visual (GoPro) cameras.
Smartplane ready to be launched for thermal mapping.
Smartplane ready to be launched for thermal mapping.

This will be evaluated later, when all data sets had been processed to orthorectified imagery and also to 3D point clouds.

An image taken with a GoPro camera from 80 meters altitude. The yellow border shows the extent of the thermal camera. Note the person walking on the road.
An image taken with a GoPro camera from 80 meters altitude. The yellow border shows the extent of the thermal camera. Note the person walking on the road.
Same as georef1_rgb.jpg, but a thermal image overlaid. White color is warm and black is cold. Note the hot area on the right side of the road which only can be seen in the thermal image.
Same as above, but a thermal image overlaid. White color is warm and black is cold. Note the hot area on the right side of the road which only can be seen in the thermal image.
An image taken with a GoPro camera from 80 meters altitude. The yellow border shows the extent of two thermal images.
An image taken with a GoPro camera from 80 meters altitude. The yellow border shows the extent of two thermal images. These images were taken about 12 hours after the fire front had passed the area.
Same as georef2_rgb.jpg, but two thermal image overlaid. White color is warm and black is cold. Note the hot spots on both side on the road only seen in the thermal images.
Same as above, but two thermal image overlaid. White color is warm and black is cold. Note the hot spots on both side on the road only seen in the thermal images.
Sunset through the smoke.
Sunset through the smoke.

Jonas Bohlin and Mattias Nyström

 

Mapping excursion to Innertavle estate

Today, the teachers and students of the Ljungberg lab, went to SLU’s new forest estate in Innertavle, close to Umeå. This forest is going to be used for teaching and research purposes. Our aim was to map it with a drone to get a high resolution orthophoto map and 3D data before the winter comes.

2015-12-04 09.55.11

Mattias is fixing with the drone  to get us in the air (in the background students making fire).

2015-12-04 08.55.31

We sprayed red crosses, which are used for ground control points (GCP) to reference the images into the national coordinate system.  These where to be positioned after the drone flight using a survey grade RTK-GNSS.

2015-12-04 13.32.38  2015-12-04 13.32.22

Unfortunately the winter came during the drone flight! Making it a very hard job to find the  GCPs both when trying to position them and in the images when doing the image processing.

But we got some result! Here is a low resolution (GSD 29 cm) orthophoto of the eastern half of the estate (click image for full size). The final products will have better geometry and resolution.

mosaic 29cm

Thanks to the students who brought BBQ lunch for us all!

/Jonas Bohlin

Gallery: