Positive Unlabelled Learning for Satellite Images’Time Series Analysis: An Application to Cereal and Forest Mapping
Applications in which researchers aim to extract a single land type from remotely sensed data are quite common in practical scenarios: extract the urban footprint to make connections with socio-economic factors; map the forest extent to subsequently retrieve biophysical variables and detect a partic...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-12-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/1/140 |
| _version_ | 1797497705177546752 |
|---|---|
| author | Johann Desloires Dino Ienco Antoine Botrel Nicolas Ranc |
| author_facet | Johann Desloires Dino Ienco Antoine Botrel Nicolas Ranc |
| author_sort | Johann Desloires |
| collection | DOAJ |
| description | Applications in which researchers aim to extract a single land type from remotely sensed data are quite common in practical scenarios: extract the urban footprint to make connections with socio-economic factors; map the forest extent to subsequently retrieve biophysical variables and detect a particular crop type to successively calibrate and deploy yield prediction models. In this scenario, the (positive) targeted class is well defined, while the negative class is difficult to describe. This one-class classification setting is also referred to as positive unlabelled learning (PUL) in the general field of machine learning. To deal with this challenging setting, when satellite image time series data are available, we propose a new framework named positive and unlabelled learning of satellite image time series (PUL-SITS). PUL-SITS involves two different stages: In the first one, a recurrent neural network autoencoder is trained to reconstruct only positive samples with the aim to higight reliable negative ones. In the second stage, both labelled and unlabelled samples are exploited in a semi-supervised manner to build the final binary classification model. To assess the quality of our approach, experiments were carried out on a real-world benchmark, namely <i>Haute-Garonne</i>, located in the southwest area of France. From this study site, we considered two different scenarios: a first one in which the process has the objective to map <i>Cereals/Oilseeds</i> cover versus the rest of the land cover classes and a second one in which the class of interest is the <i>Forest</i> land cover. The evaluation was carried out by comparing the proposed approach with recent competitors to deal with the considered positive and unlabelled learning scenarios. |
| first_indexed | 2024-03-10T03:23:03Z |
| format | Article |
| id | doaj.art-5cbd22baed43445fabc972ffadfbb6d2 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T03:23:03Z |
| publishDate | 2021-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-5cbd22baed43445fabc972ffadfbb6d22023-11-23T12:13:46ZengMDPI AGRemote Sensing2072-42922021-12-0114114010.3390/rs14010140Positive Unlabelled Learning for Satellite Images’Time Series Analysis: An Application to Cereal and Forest MappingJohann Desloires0Dino Ienco1Antoine Botrel2Nicolas Ranc3INRAE, UMR TETIS, University of Montpellier, 34000 Montpellier, FranceINRAE, UMR TETIS, University of Montpellier, 34000 Montpellier, FranceSyngenta Seeds, 31790 Saint-Sauveur, FranceSyngenta Seeds, 31790 Saint-Sauveur, FranceApplications in which researchers aim to extract a single land type from remotely sensed data are quite common in practical scenarios: extract the urban footprint to make connections with socio-economic factors; map the forest extent to subsequently retrieve biophysical variables and detect a particular crop type to successively calibrate and deploy yield prediction models. In this scenario, the (positive) targeted class is well defined, while the negative class is difficult to describe. This one-class classification setting is also referred to as positive unlabelled learning (PUL) in the general field of machine learning. To deal with this challenging setting, when satellite image time series data are available, we propose a new framework named positive and unlabelled learning of satellite image time series (PUL-SITS). PUL-SITS involves two different stages: In the first one, a recurrent neural network autoencoder is trained to reconstruct only positive samples with the aim to higight reliable negative ones. In the second stage, both labelled and unlabelled samples are exploited in a semi-supervised manner to build the final binary classification model. To assess the quality of our approach, experiments were carried out on a real-world benchmark, namely <i>Haute-Garonne</i>, located in the southwest area of France. From this study site, we considered two different scenarios: a first one in which the process has the objective to map <i>Cereals/Oilseeds</i> cover versus the rest of the land cover classes and a second one in which the class of interest is the <i>Forest</i> land cover. The evaluation was carried out by comparing the proposed approach with recent competitors to deal with the considered positive and unlabelled learning scenarios.https://www.mdpi.com/2072-4292/14/1/140land cover mappingpositive unlabelled learningsatellite image time seriesdeep learning |
| spellingShingle | Johann Desloires Dino Ienco Antoine Botrel Nicolas Ranc Positive Unlabelled Learning for Satellite Images’Time Series Analysis: An Application to Cereal and Forest Mapping Remote Sensing land cover mapping positive unlabelled learning satellite image time series deep learning |
| title | Positive Unlabelled Learning for Satellite Images’Time Series Analysis: An Application to Cereal and Forest Mapping |
| title_full | Positive Unlabelled Learning for Satellite Images’Time Series Analysis: An Application to Cereal and Forest Mapping |
| title_fullStr | Positive Unlabelled Learning for Satellite Images’Time Series Analysis: An Application to Cereal and Forest Mapping |
| title_full_unstemmed | Positive Unlabelled Learning for Satellite Images’Time Series Analysis: An Application to Cereal and Forest Mapping |
| title_short | Positive Unlabelled Learning for Satellite Images’Time Series Analysis: An Application to Cereal and Forest Mapping |
| title_sort | positive unlabelled learning for satellite images time series analysis an application to cereal and forest mapping |
| topic | land cover mapping positive unlabelled learning satellite image time series deep learning |
| url | https://www.mdpi.com/2072-4292/14/1/140 |
| work_keys_str_mv | AT johanndesloires positiveunlabelledlearningforsatelliteimagestimeseriesanalysisanapplicationtocerealandforestmapping AT dinoienco positiveunlabelledlearningforsatelliteimagestimeseriesanalysisanapplicationtocerealandforestmapping AT antoinebotrel positiveunlabelledlearningforsatelliteimagestimeseriesanalysisanapplicationtocerealandforestmapping AT nicolasranc positiveunlabelledlearningforsatelliteimagestimeseriesanalysisanapplicationtocerealandforestmapping |