I crashed my H when it sent into a tree. I don't want this to happen again.
Can RealSense sense a tree and avoid it?
Can RealSense sense a tree and avoid it?
Can Realsense avoid trees?
- Thread starter Tony2016
- Start date
From what we have seen in recent demo videos the answer is yes. But... Can it avoid thin branches sticking out in sunlight.... probably not so much as that is asking a lot.
Cheaper solution is to get the new prop guards from Yuneec.
Cheaper solution is to get the new prop guards from Yuneec.
I would expect Real Sense to be able to detect thin branches that are in the direction of flight. Prop guards aren't going to help much unless you fly into the trunk of a tree.
Yes if you look at the short vid at Yuneec it shows TH avoiding a pipe type structure. Goes up and over.
Realsense: Unfortunately technology is not able (yet) to detect thin objects like power lines, fishing line, string, tree branches smaller than your pinky finger, etc. What's worse is that sonar (which is the current OA on the Typhoon H) can not detect objects too far out or objects that are small in size. Realsense being visual is at the mercy of sight & contrast. The cameras use an advanced system of viewing what it believes to be in front of it, but it is not human and can not differentiate between what is and is not there when it comes to small objects. The sun and darkness are your enemy's in such situations. Objects in front of objects with the wrong lighting all blend together. If the object is too brightly lit, then it can't be seen. If too dark, again it can't be seen and then must rely on the existing sonar, and as we already know, the sonar is limited in range and size of object.
Propguards: Watch the many YouTube videos of Phantoms with propguards running into all types of objects (tree branches as well). Prop guards really work fantastic on a drone. The Typhoon H has six props and all it takes is one object to touch & stop one prop and that bird is coming down fast because it will throw the drone right out of alignment and into something else.
My advice to anyone who wishes to argue the above, should just wait until Realsense and the prop-guards are release and then see what the real life demos and experiences are from average users in different situations. Realsense may be the best OA on the market for hobby drones when it is released, but it will not be 100% perfect in every situation. Technology is just not yet there at this tiny price point.
Last edited:
RealSense is not at the mercy of sight and contrast because it doesn't work in the visible light spectrum, it uses infrared. Sun and darkness won't matter. Wrong lighting doesn't matter. Brightness doesn't matter. Compete darkness doesn't matter.
The sun emits a lot of IR, I use IR triggers for wireless flash guns and in bright sunlight they become unreliable. RealSense will probably need some other way to compensate for the abundance of IR bouncing around in strong sunlight.
You can't compare a IR trigger for a flash with RealSense. The former sends out very brief signal pulses from a small source that the receiver needs to pick up on, and that signal can easily get drowned out by the sun. The latter just cares about IR being emitted / reflected to the camera, whether it can decode a signal is irrelevant -- it just cares about the presence of the IR.
The only thing that RealSense should have trouble with are small objects occluding a strong IR source like the sun -- if you fly at a thin branch that is directly in front of the sun, it probably won't see it.
On the other hand, IR from the sun bouncing off an object actually helps RealSense -- that is what it is based on.
The only thing that RealSense should have trouble with are small objects occluding a strong IR source like the sun -- if you fly at a thin branch that is directly in front of the sun, it probably won't see it.
On the other hand, IR from the sun bouncing off an object actually helps RealSense -- that is what it is based on.
How will RealSense as used in the H differ from the current F200 RealSense camera module? I believe the F200 uses an IR projector to allow a structure from light 3D model to be generated, if that is the case then RealSense could very well suffer from the effects of the sun's IR swamping the scene.
The Typhoon H Pro uses the R200, not the F200. The F200 is a short-range camera, the R200 is long-range. Both have IR emitters. Here is a video of a R200 sensing and mapping an environment in realtime -- it has no trouble transitioning from indoors to outdoors and recognizes features that are probably 20 meters away (about the maximum sight distance in the video):
He is building a point cloud, a collection of 3D coordinates that can be used for navigation or obstacle avoidance. The fact that the green dots stay in the correct place as the camera moves means that the system knows their correct position and can move them relative to a change in camera position and perspective.
He is building a point cloud, a collection of 3D coordinates that can be used for navigation or obstacle avoidance. The fact that the green dots stay in the correct place as the camera moves means that the system knows their correct position and can move them relative to a change in camera position and perspective.
Last edited:
Like the F200, it relies on a IR projection and will be vulnerable to IR in sunlight, there's only one RGB camera so it needs to be able to see the IR projected patterns. If you watch the video you can see the areas in bright sunlight have much less data in the point cloud that the R200 is generating. We'll have to wait and see how it does in real world avoidance situations and how slow the H has to fly to get this to work reliably.
If it can't, it's going to be another failure of Yuneec to tell the truth in its advertising and absolutely worthless as a technology. Somehow I doubt Intel has put their stamp on something destined to fail.
From Intel's point of view it doesn't matter which drone uses their RealSense, if it works it will work on any drone that they can sell it with. If it doesn't work for Yuneec, it's not going to make much difference to Intel's fortunes, the concept will have either been proven or not. It's interesting that it's rolling out on the lower end of the prosumer market though, not the top end professional lines.
Regarding the IR emitter, skip to 9:00 in this video to see its potential effect on environment mapping. It can be useful indoors or outside at night, but its range is limited.
Where do you expect them to test the rollout of such a new and evolving product? This is exactly where I would expect to see it first.
Intel isn't just selling Yuneec the R200 cameras, they invested US$ 60 million in Yuneec and are helping them integrate the technology. Given that Intel created the technology, Yuneec has the backing and support of the people who know it best -- I don't expect a half-a$$ed implementation.
Oh yes it does, it "Is" a camera, two actually !
BUT IR is much less susceptible to the visible light portion of the spectrum so it is much less affected in darker lighting.
