Garmin LIDAR v4 거리 오차에 대한 문제

Sensors Distance
1

Hello. I found a critical problem while measuring objects using LiDAR.

I measured an object positioned parallel to the LiDAR at 10cm intervals from 50cm to 120cm.

image
image1368×870 8.7 KB

I confirmed that the measured data showed LiDAR values up to 4.6cm shorter than the actual distance, and the error increased linearly with distance.

I had been looking at this part of the specification and thought it meant absolute accuracy, so I thought my experimental setup was wrong, but then I realized it was precision. Absolute accuracy is not specified anywhere in the specifications.

image
image467×123 13.7 KB

My question is, what is the distance error in centimeters that your company can guarantee, plus or minus?

I need to know this so I can stop investigating the experimental environment and just calibrate and move on.

Please reply quickly.

2

Keep in mind that Sparkfun is a distributor of the Garmin LIDAR.

The Garmin company should be contacted with questions about measurement accuracy, as opposed to measurement precision. Absolute accuracy does not appear to be stated in the LIDAR V4 LED technical manual.

I would not expect the LED version to be as accurate or as precise as a laser version, but for the LIDAR-Lite v3HP, the following is stated, and your observations with the V4 LED are consistent with these claims for the v3HP laser version.

Capture
Capture621×237 24.1 KB

3

As above; Precision/resolution = how repeatable readings are at a fixed distance (spread/noise)

Accuracy = how close the reading is to the true distance (systematic bias, this is what you’re seeing).

Garmin doesn’t publish a number for it on the v4 LED as returned signal strength falls off with distance (even for a fixed target/reflectance), and the v4 LED’s correlation/threshold-based time-of-flight detection has a signal-strength-dependent bias (“walk error”). Reflectance, target size, angle, and ambient light all feed into that same signal-strength variable, which is why Garmin can’t commit to one absolute-accuracy number across use cases…it’s condition-dependent, not a fixed device spec. If you changed the lighting or object’s color/reflectance you’d see slightly different results

The good news is if you’re going to measure the same object in similar lighting you can physically measure and implement an offset (you’ve basically already done the leg-work here) in code that is more realistic and yields improved results

Also like mentioned above the v3HP (laser) version does list accuracy as a spec, if you require this it might be the better option