Seasonal leaf area data of apple trees (3 years) read by reference LiDAR and lowcost RGB-D sensors

Apple trees were scanned with two LiDAR sensors (SICK LMS511 and Pepperl & Fuchs). Both sensors provide 3D point clouds of each tree to gain geometric information. Such geometric information provide the gold standard for leaf area analysis at present. Additionally the SICK LMS511 provides the return signal strength intensity at 905 nm and the P&F the RSSI measured at 660 nm. These intensity values at specific wavelengths allow a dual wavelength 3D point cloud analysis. Such wavelength-dependant analysis can provide information on the fruit ripeness based on the decrease of fruit chlorophyll content.

The third sensor captured is a lowcost sensor: Intel RealSense 435. The RealSense provides RGB and depth information. For each tree 20 RGB images and the depth information is captured.

This example dataset captures a part of a bigger dataset, representing an example of one tree measured from two sides (right R, left L) for three years. The example data were selected randomly (27.09.2021, 07.09.2022, 12.06.2023). 

The entire dataset available at ATB captures the data (examplarily shown here for 1 tree in three years) for ca. 80 trees on each of the following dates, summing up to approx. 1400 trees:

Plot
Year
Date
LiDAR (LMS511 + P&F) (GB)
Intel RealSense 435 (GB)

MQT
2021
06.07.2021
7.9
12.7

MQT
2021
12.07.2021
13.2
50.6

MQT
2021
24.08.2021
15.0
48.3

MQT
2021
06.09.2021
15.4
50.1

MQT
2021
27.09.2021
22.8
155.0

MQT
2022
11.05.2022
24.0
48.0

MQT
2022
29.06.2022
38.5
207.0

MQT
2022
12.07.2022
12.8
131.0

MQT
2022
17.08.2022
15.7
79.0

MQT
2022
04.08.2022
17.0
164.0

MQT
2022
07.09.2022
20.4
67.7

MQT
2023
22.08.2023
15.6
529.0

MQT
2023
06.09.2023
84.0
167.0

MQT
2023
03.11.2023
7.6
103.0

MQT
2023
11.04.2024
9.9
68.5

MQT
2024
24.05.2024
35.1
529.0

MQT
2024
03.06.2024
36.9
448.0

MQT
2024
09.07.2024
13.6
451.0

Such dataset allows to follow the growth of leaf area (or leaf area index) over the seasonal. The dataset provided opportunities

to test various data analysis methods for leaf area estimation, based on manual reference data, which were obtained by defoliation method.

to compare the expensive LiDAR with lowcost sensor unit.

developing growth curves to interpolate the leaf area for comparing with satellite data, when available.

to run various approaches to fuse the data and, e.g. perform dual wavelength analysis in 3D.

to test data processing steps for fruit ripeness analysis, based on manual fruit size readings.

test own physiological growth models against this open access dataset.

The dataset is given in following formats:

csv files contain one 3D pointcloud of each individual trees measured from two sides, at each measurement date.

csv files contain Sentinel 2 data of the plot. The file name indicates the satellite overpass.

txt files contain depth information by RGB-D sensor measured twenty times of each individual trees measured from two sides, at each measurement date.

bmp files contain RGB images by RGB-D sensor measured 20 times of each individual trees measured from two sides, at each measurement date.

xlsx files capture manual reference analyses of leaf area per tree (n = 37) and fruit (n > 500).