Sensor suite suggestions

ArduCopter
, , ,
1

Our current drone system is designed to spray pesticides in fields while navigating using a rangefinder to maintain a consistent height of 3 meters above the crop. Typically, this setup works well, as most crop canopies are uniform, and distinguishing between reflections from the canopy and the ground isn’t crucial.

However, when dealing with certain crops like sugarcane, which can have variable canopy heights ranging from 4 to 8 meters, this presents a challenge. In such cases, reflections in the rangefinder can originate from either the ground or the canopy, depending on the density of the crop.

To address this issue, I’m exploring various sensor options that can accurately differentiate between the canopy and the ground. Here are my thoughts on some available sensors:

Lidar: While Lidar could provide a reliable estimate of the sugarcane canopy’s top, its performance might be affected by environmental factors such as spray (creating foggy conditions). Additionally, implementing an IP-protected rotating Lidar could be a pain.

Radar: Radar sensors offer the potential for complex signal processing to distinguish reflections from the canopy and the ground. My current hypothesis suggests that the reflection could originate from the middle of the crop, necessitating sophisticated algorithms to interpret radar signals accurately.

Stereocameras: would be blurred by spray

Looking forward to the thoughts from the AP community for this

2

For sugarcane what do you want to happen? Do you want each area sprayed exactly 3 ft above or you just want to be 3 ft above the plants?

I don’t know sugarcane well enough but this seems more like a software problem vs. hardware. My intuition says one rangefinder should be sufficient if you do post-processing. Have you done a run over a field to see what data you get? I wouldn’t be surprised if you could just be dumb and take the 99th percentile height + 3 feet and call it a day.

3

Data collection is in progress. My idea was to interface all the sensors that may be proposed and get the data. Ideally i would like to fly 3m above the canopy, using a rangefinder to follow the terrain/crop.
If reflection is always from ground (I can roughly estimate crop canopy’s height and decide my flying altitude)
If reflection is always from sugarcane canopy then no worries.
However, in our case reflection point varies based on density of sugarcane (comes from top of canopy for dense fields, and from ground in case sugarcane is scanty). This distribution is generally random in the field.

We want to estimate the height in real time. (while doing terrain following, hence post processing is not the goto plan). But for POC I can think of postprocessing frameworks too.