CONSUMER DRONESComing soon
A small consumer drone can produce a useful orthomosaic, surface model and 3D model. Stream Capture reads the camera data from the photos, treats the drone's ordinary GPS as the weak position it is, and can correct for the rolling shutter these cameras use. What a consumer drone cannot give you is survey-grade absolute position without ground control points.
Stream Capture is in development and not released yet. At launch it will come in the same download as its sister program Stream Terrain; today's download contains Stream Terrain.
A consumer drone writes an ordinary GPS position into each photo, good to metres rather than centimetres, and often reads its sensor line by line with an electronic shutter. The map it makes can be internally consistent and still sit several metres off on the ground, and the shutter can bend straight lines on a moving camera.
Stream Capture handles both honestly. The GPS position is used only as a soft prior, and the report measures the result against any control or check points you supply, so you see the error instead of assuming it away.
JPEG, TIFF and PNG photos from any camera make are read with their EXIF and XMP metadata. A sensor table keyed on make and model converts the focal length in millimetres to pixels, which avoids the roughly 30% focal error that one assumed sensor size gives on cameras of a different size.
When no accuracy is reported, positions weigh in at 3 m horizontal and 5 m vertical and camera centres stay free, so a poor GPS fix cannot drag a good photo network out of shape.
Focal length and distortion are solved from the photos and reported with their uncertainty. When the lens data in the photos is reliable, a self-calibrated focal length that strays more than 2% from it is not adopted.
For sensors read line by line, the readout time is estimated in the adjustment and applied only if it improves the alignment error by at least 2% and is determined with confidence. Dense matching and the orthophoto assume a global shutter, and the option is not used with split adjustment.
Add ground control points, or at least check points, to bring a consumer-GPS project onto the ground. The GCP editor, automatic target marking and the independent check-point error work the same as for any project.
Fast Draft shows holes and unaligned areas quickly, and the overlap map in the Quality Report shows which parts of the site too few photos saw.
Yes. The photos are processed like any others, and the result can be a good orthomosaic and surface model. Absolute position is limited by the drone's consumer GPS, so add ground control points when position accuracy matters.
It depends on the camera, the overlap, the height and the control. Stream Capture does not promise a number: it measures your result at the control and check points you provide and reports the error.
If the map only has to look right and be internally consistent, no. If it has to match coordinates on the ground, yes: consumer GPS is good to metres, and control points are what fix the position and scale in your coordinate system.
Many consumer cameras read the sensor line by line, so each line is captured at a slightly different camera position when the drone moves. Stream Capture can estimate and correct this in the alignment; the option is used only when it measurably improves the result.
No. Video input is not supported. Stream Capture reads still photos in JPEG, TIFF and PNG.
At least three overlapping photos are required, and a real map needs far more. Plan generous overlap; the health check estimates the overlap of your photos and the overlap map in the Quality Report shows weak areas.
Stream Capture is in development and not released yet. At launch it will come in the same download as its sister program Stream Terrain; today's download contains Stream Terrain.
About Stream Capture