Thanks for the suggestion. After long night of trying I still have to disappoint you guys:
1, PX4Flow still gives spiking results. Following every step from the instruction, we are still getting extreme values from the PX4flow at 10e5. We have used the right firmware. Our PX4Flow is bought from https://www.unmannedtechshop.co.uk/px4flow-smart-camera-optical-flow-sensor/?utm_source=Unmanned+Tech+Customers&utm_campaign=cfe2fcd45c-EMAIL_CAMPAIGN_2017_04_27&utm_medium=email&utm_term=0_ecf952454c-cfe2fcd45c-213082101&mc_cid=cfe2fcd45c&mc_eid=ff2c63e173, which is slightly different from the explained one and comes without screws on the lens.
A closer look at the Log one can see that the PX4Flow Bodyrate data is very strange.
The body rate has huge spikes that caused the spiking flow measurement. They appear quite randomly.
This could be related to power issue. Since this has happended with other user already http://discuss.ardupilot.org/t/px4flow-spikes-on-data-output/14963/60, and that it works correctly with USB connection, I would claim this to be a hardware failure related rapsberry pi 3 or Erle Firmware. Up to now, PX4flow doesn't seem to be compatible with Erle brain 3.
The LidarLitev3 gives reading in Copter v3.4rc2, but can't seem to be intergrated for altitude estimation, if I understood correctly. The
EK2_ALT_SOURCE says that additional optical flow sensors are required. I don't know whether it helps if we can't use PX4flow.