Matlab implementation for fruit detection in 3D LiDAR point clouds.

Introduction

This project is a matlab implementation for fruit detection in 3D point clouds acquired with LiDAR sensor Velodyne VLP-16 (Velodyne LIDAR Inc., San Jose, CA, USA).

This implementation was used to evaluate the LFuji-air dataset, which contains 3D LiDAR data of 11Fuji apple trees with the corresponding fruit position annotations. Find more information in:

• Fruit detection, yield prediction and canopy geometric characterization using LiDAR with forced air flow [1]

Preparation

First of all, clone the code

git clone https://github.com/ GRAP-UdL-AT/fruit_detection_in_LiDAR_pointClouds.git

Then, create a folder named “data” in the same directory where the code were saved. Inside the /data folder, save the ground truth and point cloud data (“AllTrees_Groundtruth” and “AllTrees_pcloud”) available at LFuji-air dataset.

Pre-requisites

• MATLAB R2018 (we have not tested it in other matlab versions)
• Computer Vision System Toolbox
• Statistics and Machine Learning Toolbox

Data Preparation

• LFuji-air dataset: Save the LFuji-air dataset in /data folder.

Cross-vailidation (fruit detection)

Open the matlab file main_CrossVal_Velodyne_fruit_detection.m and set the following parameters:

directory = $”code_directory”$;    %Write the directory where the code and the /data folder are placed.
Trials2eval = ${trials_to_evaluate}$;    %List the trials to evaluate

example:

directory = 'F:\fruit_detection\vel_air';  
Trials2eval = {'H1_n_E_O','H1_n_E','H1_n_O','H1_H2_n_E_O','H1_n_af_E_O'}; 

Execute the file main_CrossVal_Velodyne_fruit_detection.m.

Train (fruit detection)

Open the matlab file main_Velodyne_fruit_detection.m and set the following parameters:

directory = $”code_directory”$;   %Write the directory where the code and the /data folder are placed.
pcDiectory_txt = strcat(directory, $”training_data_folder”$ );   %Write the name of the training data folder.
train = $logical_number$; %Set this parameter to 1 for training the svm models.

example:

directory = 'F:\fruit_detection\vel_air';  
pcDiectory_txt = strcat(directory, '\data\TrainingData\');
train = 1;

Execute the file main_Velodyne_fruit_detection.m.

Test (fruit detection)

Open the matlab file main_Velodyne_fruit_detection.m and set the following parameters:

directory = $”code_directory”$;   %Write the directory where the code and the /data folder are placed.
pcDiectory_txt = strcat(directory, $”test_data_folder”$ );   %Write the name of the test data folder.
train = $logical_number$; %Set this parameter to 0 to test data using a previously trained svm models.

example:

directory = 'F:\fruit_detection\vel_air';  
pcDiectory_txt = strcat(directory, '\data\TestData\');
train = 0;

Execute the file main_Velodyne_fruit_detection.m.

Canopy geomtry characterization

In [1], the LFuji-air dataset is used to evaluate the fruit detection performance, but also to compute canopy geometrical measurements such as mean height, mean width, canopy contour, mean canopy cross section and leave area. To compute this canopy geometrical parameters from a 3D LiDAR point cloud, do the following:

Open the matlab file main_Velodyne_LA_meanShape.m and set the following parameters:

directory = $”code_directory”$;    % Write the directory where the code and the /data folder are placed.
Trials2eval = ${trials_to_evaluate}$;    %List the trials to evaluate

example:

directory = 'F:\fruit_detection\vel_air';  
Trials2eval = {'H1_n_E_O','H1_n_E','H1_n_O','H1_H2_n_E_O','H1_n_af_E_O'}; 

Execute the file main_Velodyne_LA_meanShape.m.

Authorship

This project is contributed by GRAP-UdL-AT.

Please contact authors to report bugs @ [email protected]

Citation

If you find this implementation or the analysis conducted in our report helpful, please consider citing:

@article{Gene-Mola2019,
    Author = {Gen{\'e}-Mola, Jordi and Gregorio, Eduard and Auat Cheein, Fernando and Guevara, Javier and Llorens, Jordi and Sanz-Cortiella, Ricardo and Escol{\`a}, Alexandre and Rosell-Polo, Joan R},
    Title = {Fruit detection, yield prediction and canopy geometric characterization using LiDAR with forced air flow},
    Journal = {Submitted},
    Year = {2019}
} 

References

[1] Gené-Mola J, Gregorio E, Auat Cheein F, Guevara J, Llorens J, Sanz-Cortiella R, Escolà A, Rosell-Polo JR. 2020. Fruit detection, yield prediction and canopy geometric characterization using LiDAR with forced air flow. Computers and Electronics in Agriculture, 168 (2020), 105121. DOI: 10.1016/j.compag.2019.105121.

Acknowledgements

This work was partly funded by the Spanish Ministry of Science, Innovation and Universities (grant RTI2018-094222-B-I00[PAgFRUIT project] by MCIN/AEI/10.13039/501100011033 and by “ERDF, a way of making Europe”, by the European Union).