Processing Survey3 Camera Images

The below information explains the best workflow to post-process images from the Survey3 cameras.

Capturing Images

First, off, we recommend you read the manual for the Survey3 from cover to cover. We know manuals may be boring, but it's not that long a read and will answer a lot of the questions you may have.

The Survey3 camera will come with the best default settings that we expect to work for the majority of our customers, so there's a good chance you will not need to change them. It is not recommended to adjust any camera photo settings (shutter speed, ISO, white balance, etc) unless you know what you're doing and have our Reflectance Calibration Target. We have already set the default values so that you should not get over-exposed pixel values, but this likely makes the images darker than you're used to. Darker than normal images are fine for processing as it reduces the chances that a pixel would be over exposed, and thus lose that pixel data. If you have questions about adjusting the settings, feel free to reach out to us for help, but when in doubt just use the defaults.

You no longer need to manually set the camera time, simply connect the GPS receiver and get a lock to correct for any time drift during use. If you are using the GPS receiver (recommended) then the time will update every time it gets a good lock.

Even if you are not planning on using the included GPS receiver for geo-tagging the images, we recommend you connect it when you first get the camera. With the camera powered on and the GPS installed walk around outside with the GPS receiver pointing upward (side opposite the adhesive) until you hear 6 beeps, signaling the camera has a good GPS lock. This lock will then update the time on the camera to GPS time. You can then remove the GPS receiver if you don't plan on using it.

The Survey3 cameras can be told to take a photo either by its default internal timer (intervalometer) or using an external PWM signal received via the HDMI port

The time between the timer triggered shots is a total of the camera's Interval setting value plus the time it takes to save the image to the camera's memory card. When we say the fastest the JPG mode can be captured every 1.5s, this time includes the 0.5s Interval settings time plus about a second to save the image. It's rare that you'd want to increase the Interval time, unless you do indeed want to capture less images. The 1.5s is also the average the timer mode will trigger at, as it will vary depending on camera settings, SD card memory speed and internal camera processes.

For the most reliable triggering we highly recommend you use the internal timer, as using the PWM signal can produce issues with missed images due to many variables within the camera. For instance, the model of the SD card you use, and how quickly it can save the image can affect how quickly the camera is ready again to capture the next photo. The memory cards we sell are what we used to determine the fastest capture rates, so using slower cards (meaning any other brand or model card) will likely trigger slower. Triggering the camera faster than it can save the previous image will cause the camera not to be ready and thus you will miss that new image. You can use our camera flight calculator at the bottom of the page here to see the fastest that a camera has to be triggered in order to provide the overlap you require for your survey. Pay attention to the "Photo Interval" line towards the bottom.

If you are triggering the camera via PWM make sure to set the Interval setting to OFF. 

The camera will default to saving images only as JPG. In order to capture larger contrast in your images, along with not having any compression and gamma applied, we recommend changing to the RAW+JPG mode. Setting it to RAW+JPG mode takes longer for the camera to save the image, so the time between triggers will be longer. In our MAPIR Camera Control (MCC) application you can easily convert the RAW to TIFF to JPGs if your stitching software requires JPGs (such as DroneDeploy). The less compression (more RAW) your images are the easier they are to stitch as well, so use TIFFs whenever possible.

Survey mission planning is based on your aerial platform and what flight controller it is using. This video shows how to setup a mission with a Pixhawk-based drone and the free Android mission planning application Tower.

Prepare Images in MAPIR Camera Control Application

Once you capture your images you may want to prepare them prior to stitching them into a 2D orthomosaic image or generating a 3D model, that's where our MAPIR Camera Control (MCC) application is used.

If you captured the images in RAW+JPG mode, then use the Process tab of MCC to convert to TIFFs. During processing the full image metadata including the GPS capture location (if using the included GPS receiver) will be copied from the JPG to the TIFF.

If you are going to use ortho-mosaic generation software like Pix4D and Agisoft Photoscan then you will want to stitch the ortho first and then proceed to calibration below. Do not calibrate and then stitch as the results will not be satisfactory (poor automatic tie points). It's also not recommended to use the the calibration and index steps of these software packages either as they are lacking in both features and accuracy (only single reflectance target versus our multiple target calibration).

Next you will want to calibrate the images (not necessary with a RGB Visible Light model) using the Calibrate tab of MCC. This step will ask you for an image of our Reflectance Calibration Target if you captured one before your survey (recommended), or if you don't have one you can use the default reflectance values we provide (created from an image taken during full sun). The images will be calibrated with each pixel representing the reflectance value and not a color value, so don't be concerned if the images seem too dark. You can also choose to convert the TIFF files to JPG if your ortho-mosaic software does not support TIFFs (such as DroneDeploy, MapsMadeEasy, etc).

Creating the Ortho-mosaic

An ortho-mosaic is a single stitched image containing the many individual photos taken during your survey. Your results will vary though based on the capabilities of the software used to create the ortho image.

Software that is not able to use an image's GPS data to assist stitching will struggle to create ortho-mosaics of complex subject matter. Complex subject matter is typically captured when flying over areas where the images will look very similar to each other, such as uniform canopy agricultural fields. Software that doesn't use the GPS data will also have no way to locate the final stitched ortho image onto a map (like Google Earth) because there will be no reference GPS information. We strongly recommend you do not use software that cannot take advantage of the GPS location in the survey images.

There are two options to process your images with GPS information: cloud-based services that do the processing for you or stand-alone programs you run yourself.

Cloud-based packages such as DroneDeploy, MapsMadeEasy all require you to upload your images to their site and then notify you when the processing is completed. These services vary in how they charge you but the majority have a minimum monthly fee around $100. They also vary in the final results you'll get from their services, and typically the more expensive "professional" packages provide more outputs with higher cost. Output examples include geo-reference images, DSM maps, KML files and NDVI color-graded images. These cloud-based services often do not typically support RAW/TIFF files so make sure to convert the input image to JPGs using the calibration step in MCC mentioned above. They also do not support aligning image datasets from multiple cameras with each other, so you cannot layer the orthomosaic images into a composite image.

Two of the top packages which support aligning the single band cameras with each other are Pix4D and Agisoft Photoscan. These programs use structure from motion (SfM) methods, and are more commonly known as "point cloud" software. They are called this because they look at each and every point, or more specifically each pixel and match up the images. While they can stitch some image sets this way without GPS information, it will greatly reduce your processing time and increase your success rate if you do use geo-referenced photos.  The importance of aligning the cameras with each other is that you can then process the ortho layers in a raster/index calculator for the indice (ie NDVI, ENDVI, GNDVI, OSAVI, RDVI, SAVI, etc) you require. Both Pix4D and Photoscan have raster calculators included but since the images won't be calibrated until after ortho generation we recommend calibrating in MCC and then using the Raster and Lut options in the Viewer tab of MCC. Since these programs create a point cloud they can also output .obj, .mtl and .jpeg texture files to be used in 3D model viewers like p3d and Sketchfab.  These powerful point cloud software packages typically cost about 3x more than the cloud-based services but the additional cost should be easily amortized over time due to the ability to align the single band images to calculate the index of your choice. Some cloud-based solutions are also beginning to support SfM such as DroneDeploy and Pix4D's own Cloud platform, but typically you still can't process multiple datasets with each other.