Efficient Multi-Resolution Fusion for Remote Sensing Data with Label Uncertainty
Hersh Vakharia and Xiaoxiao Du
[arXiv] [IEEEXplore]
This code uses MATLAB Statistics and Machine Learning Toolbox, MATLAB Optimization Toolbox and MATLAB Parallel Computing Toolbox.
Run demo_main.m in MATLAB.
The MIMRF-BFM Algorithm runs using the following function:
[measure, initialMeasure, Analysis] = learnCIMeasure_bfm_multires(Bags, Labels, Parameters)
The Bags input is a 1xNumTrainBags cell array. Inside each cell, NumPntsInBag x nSources cell. Inside each cell, the "collection" of all possible combinations generated from the multi-resolution data set.
The Labels input is a 1xNumTrainBags double vector that takes values of "1" and "0" for two-class classification problems - Training labels for each bag.
The parameters can be set in the following functions:
[Parameters] = learnCIMeasureParams()
The parameters is a MATLAB structure with the following fields:
- eta: percentage of time to make small-scale mutation
- analysis: if = "1", save all intermediate results
- exaustiveSearchThresh: count threshold for number of repeated samples
- fitnessUpdateThresh: count threshold for number of times new BFM samples do not improve over past iterations
This source code is licensed under the license found in the LICENSE file in the root directory of this source tree.
This product is Copyright (c) 2023 H. Vakharia and X. Du. All rights reserved.
If you use the MIMRF-BFM multi-resolution fusion algorithm, please cite the following reference using the following BibTeX entries.
@INPROCEEDINGS{10282851,
author={Vakharia, Hersh and Du, Xiaoxiao},
booktitle={IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium},
title={Efficient Multi-Resolution Fusion for Remote Sensing Data with Label Uncertainty},
year={2023},
volume={},
number={},
pages={6326-6329},
doi={10.1109/IGARSS52108.2023.10282851}
}