The point cloud obtained with OAK-FFC-ToF-33D's tof is noisy, so is this a problem with the tof sensor itself? On the way is the point cloud obtained by the tof, the red line represents the ground, as can be seen in the figure, there are also some clouds below the ground (in fact, this is obviously impossible), so to doubt whether this is the tof sensor itself？

    zenglu
    What are the settings (phase unwrapping and filtering) you are applying in the script? Can you show the color image as well?

    Thanks,
    Jaka

        jakaskerl

        tofConfigA.enableOpticalCorrection = true;

        tofConfigA.enablePhaseShuffleTemporalFilter = true;

        tofConfigA.phaseUnwrappingLevel = 4;

        tofConfigA.phaseUnwrapErrorThreshold = 300;

        tofConfigA.enableTemperatureCorrection = true;

            zenglu
            Difficult to say without seeing the base image. I'd say keep unwrapping level as minimal as you can at short distances since it introduces a lot of noise.

            Thanks,
            Jaka

                jakaskerl

                When I set the phaseUnwrappingLevel to 2 and 1 respectively, there are a lot of noise in front of tof (about 0 ~ 50cm) which is very close to TOF (and I am sure there are no obstacles in this place, so I think it is noise), and the smaller the phaseUnwrappingLevel is, The worse the noise. Why is that? The original depth map below is the depth map at phaseUnwrappingLevel=1.

                    zenglu This is the original depth map and point cloud of the interior environment scanned by tof inside the cylinder tube

                      zenglu
                      Yeah, that is noise. If you know no object is present close to the device, consider adding a threshold filter to discard the values. Or perhaps a median filter (setMedianFilter) with a large enough kernel should work as well..

                      Thanks,
                      Jaka