Could Ingenuity use the shadow of its legs for navigation to fly autonomously to the sample cache depot location?

Question

It's summertime now in Jezero crater, and according to the HORIZONS Web-Interface the subsolar latitude for Mars varies between both the end of May and November 2021 from 18.5⁰ N to ~25.5⁰ N and back to 18.5⁰ N again.
With Jezero crater being at 18.38⁰ N latitude this means the path of the Sun through the sky there in that period will be almost exactly from east to west.

Credits: NASA/JPL

On the image we can see that the legs are placed 90⁰ from each other and that the Sun gives a clear shadow of those legs on the surface. And when orientated in the right direction, in the early morning or late afternoon one of the legs could have a small shadow line in the field-of-view of the navigation camera. (see Fig. 11 from Mars Helicopter Technology Demonstrator)

Could an image of such a shadow line not be processed every day during takeoff in such a way that together with the onboard IMU a chosen flight direction could be maintained ?
If for instance, during takeoff in the morning the shadow of the right rear leg would cross the field-of-view of the downward-looking camera, the navigation unit could "know" that the front of the helicopter is directed to the northwest.

So the ultimate question is: "Could Ingenuity's science team change the navigation software in such a way that with processing the images taken by the navigation camera capturing a shadow line of one of Ingenuity's legs, together with the data from the IMU, the helicopter will be able to fly autonomously in one direction, after the test campaign has ended ? "

Answer 1

Probably not, at least with a downward looking camera. It is certainly possible with dedicated hardware.

With a known flat surface and a sequence of images around local Noon it is possible to find north and with a count of days and some math allowing for inclination it is possible to determine Latitude. Dealing with the leg geometry, the fact that the surface landed on will not be perfectly level makes this a lot more complicated.

The Accelerometers would allow the attitude of the camera to vertical to be determined, but this may not be the angle of the surface the shadows are being cast on due to one or more legs being in sand, or local lumps/rocks under the camera. A one degree combined error from the surface geometry and sensor precision would be around 60km.

With sufficient image processing to use shadow movement to 3D map the area under the camera this could probably be handled but would not expect it to be possible with the power and CPU budget onboard.

And none of this solves the problem of longitude, which normally requires either an accurate clock, or precise angular measurement of known stars. Unsure on the precision of the onboard clock, but given the temperature range of operation would not expect it to be particularly high. At the equator a one second error in time keeping from the known start position would give around 240 meters of error in measure longditude, assuming actual local noon could be determined (see above).

It appears Inginuity carries an upwards facing sun sensor. Wikipedia links to a thesis giving three and a half degree errors, which would not be very useful for multi day navigation, but is also not from the as flown sensor.

The terrestrial autonomous stellar navigation systems do indicate that relatively small payloads of dedicated hardware can provide a navigation fix, which might provide a useful fail safe to bring things like rovers supporting a manned mission back close enough for retrieval after some types of radio failure, in particular if vehicle features a moveable camera using it as a sextant becomes possible side stepping a number of error sources mentioned above.