Is Boteye fit for Greenhouse Scene?

[zhanglei2015yy]

I test boteye in a greenhouse for Tomato. There are about 100 rows crops. There's a natural dirt road between each row,on which my robot run. You can refer the pic below.

It seems like that ,before BA, the trajectory is consistent with the actual.

But after BA, the trajectory have large deformation. The roundnd trip trajectory is not consistent.

So My question is:

1. Is boteye fit for greenhouse scene?because the crop will grow day by day.
2. Why the trajectory have large deformation before & after BA? 3.I test boteye in a greenhouse for Tomato. There are about 100 rows crops. There's a natural dirt road between each row,on which my robot run. You can refer the pic below.

It seems like that ,before BA, the trajectory is consistent with the actual.

But after BA, the trajectory have large deformation. The roundnd trip trajectory is not consistent.

So My question is:

1. Is boteye fit for greenhouse scene, because the crop will grow significantly week by week?
2. Why the trajectory have large deformation before & after BA?
3. Will sunlight at different time cause diffrenet feature points in the same scene filled with crops?

[mingyux]

This is a very interesting scene that we have never encountered before. I'll try to answer all three questions at once. The VIO part will work in your environment for sure, but we're not sure about how well the loop detection work in the greenhouse scene. It is possible that a false loop detection causing the large deformation after BA. Also, the relocalization can be tricky in your environment as well. In addition to the unstructured scene, the changes of light and crops over time will make the relocalization harder and eventually fail. Our internal version of boteye may have some features that work in your scene. If interested, please contact us at ext_roboticvision@baidu.com.