Pub Date : 2020-03-01DOI: 10.1109/LAGIRS48042.2020.9165562
W. Ramirez, P. Achanccaray, Leonardo A. F. Mendoza, Marco Aurélio Cavalcanti Pacheco
The presence of weeds in agricultural crops has been one of the problems of greatest interest in recent years as they consume natural resources and negatively affect the agricultural process. For this purpose, a model has been implemented to segment weed in aerial images. The proposed model relies on DeepLabv3 architecture trained upon patches extracted from high-resolution aerial imagery. The dataset employed consisted in 5 high-resolution images that describes a sugar beet agricultural field in Germany. SegNet and U-Net architectures were selected for comparison purposes. Our results demonstrate that balancing of data, together with a greater spatial context leads better results with DeepLabv3 achieving up to 0.89 and 0.81 in terms of AUC and F1-score, respectively.
{"title":"Deep Convolutional Neural Networks for Weed Detection in Agricultural Crops Using Optical Aerial Images","authors":"W. Ramirez, P. Achanccaray, Leonardo A. F. Mendoza, Marco Aurélio Cavalcanti Pacheco","doi":"10.1109/LAGIRS48042.2020.9165562","DOIUrl":"https://doi.org/10.1109/LAGIRS48042.2020.9165562","url":null,"abstract":"The presence of weeds in agricultural crops has been one of the problems of greatest interest in recent years as they consume natural resources and negatively affect the agricultural process. For this purpose, a model has been implemented to segment weed in aerial images. The proposed model relies on DeepLabv3 architecture trained upon patches extracted from high-resolution aerial imagery. The dataset employed consisted in 5 high-resolution images that describes a sugar beet agricultural field in Germany. SegNet and U-Net architectures were selected for comparison purposes. Our results demonstrate that balancing of data, together with a greater spatial context leads better results with DeepLabv3 achieving up to 0.89 and 0.81 in terms of AUC and F1-score, respectively.","PeriodicalId":111863,"journal":{"name":"2020 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128850213","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-03-01DOI: 10.1109/LAGIRS48042.2020.9165678
D. Castillo, A. Russell, V. Caquilpan, S. Elgueta
High-Andean wetlands from northern Chile are considered worldwide biodiversity hot spots, however, they are subdued to high anthropic pressure. The monitoring of state variables, such as vegetation, allows to know the ecosystem’s global condition, which could be assessed by the analysis of spectral vegetation indices. The main goal of this paper was to detect changes in the high-Andean wetland vegetation, with remote sensing tools, to focalize surveillance efforts and the use of resources from environmental agencies. NDVI time series were constructed spanning from 1986 to 2019 based on Landsat data, which were analyzed based on the vegetation change detection using BFAST Monitor method. Detected changes were categorized to highlight certain types of changes that were considered more relevant. Wetlands were separated in two rankings (A and B) based on detected changes and territorial context. From 5,622 wetlands, 81 were categorized into group A and 510 into group B. One affected wetland was used as study case to assess the method’s efficiency, being able to detect changes and assign a relative importance to the case. It is shown that the proposed method has the capacity to detect vegetation degradation processes in high-Andean wetlands and could improve in the efficiency and effectiveness of the environmental agencies control labors over these ecosystems.
{"title":"Identification Of Affected High-Altitude Wetlands In The North Chile Using Large Landsat Time Series","authors":"D. Castillo, A. Russell, V. Caquilpan, S. Elgueta","doi":"10.1109/LAGIRS48042.2020.9165678","DOIUrl":"https://doi.org/10.1109/LAGIRS48042.2020.9165678","url":null,"abstract":"High-Andean wetlands from northern Chile are considered worldwide biodiversity hot spots, however, they are subdued to high anthropic pressure. The monitoring of state variables, such as vegetation, allows to know the ecosystem’s global condition, which could be assessed by the analysis of spectral vegetation indices. The main goal of this paper was to detect changes in the high-Andean wetland vegetation, with remote sensing tools, to focalize surveillance efforts and the use of resources from environmental agencies. NDVI time series were constructed spanning from 1986 to 2019 based on Landsat data, which were analyzed based on the vegetation change detection using BFAST Monitor method. Detected changes were categorized to highlight certain types of changes that were considered more relevant. Wetlands were separated in two rankings (A and B) based on detected changes and territorial context. From 5,622 wetlands, 81 were categorized into group A and 510 into group B. One affected wetland was used as study case to assess the method’s efficiency, being able to detect changes and assign a relative importance to the case. It is shown that the proposed method has the capacity to detect vegetation degradation processes in high-Andean wetlands and could improve in the efficiency and effectiveness of the environmental agencies control labors over these ecosystems.","PeriodicalId":111863,"journal":{"name":"2020 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115058072","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-03-01DOI: 10.1109/LAGIRS48042.2020.9165571
J. Cañete, C. Cárdenas, M. Frangópulos, X. Aguilar, J. Díaz-Ochoa
Zooplankton aggregation, hydrographic and remote sensing data were employed to relate the spatial dynamics of neustonic communities with chlorophyll a (Chl a) and suspended organic matter (SOM) at a spatial mesoscale (10 to 1000 km) in the southem Chilean fjords system along Magellan Strait, Chile (CIMAR 16: October/November 2010 and CIMAR 25; September/October 2019) in order to identify oceanographic process producing aggregation of neuston. Preliminary evidence of CIMAR 25 shows significant concentrations of Chl a and SOM around Dawson Island (DI), Magellan Strait. During CIMAR 16 important aggregation of specific neustonic taxa (copepodites of Microsetella rosea, larvae of the polychaete Polygordius sp and cyphonautes of the bryozoan Membranipora isabelleana) was observed around DI, Magellan Strait. Satelital images in the area of CIMAR 16 provide evidence of important aggregation of chlorophyll a/SOM around DI. CIMAR Cimar 25 showed that the Chl a and SOM aggregation around DI is recurrent and could to explain the high concentration of neuston around this island to spite of mesotrophic conditions. Remote sensing in this study area provides a tool to understanding oceanographic and topographic factors that potentially regulate the abundance and spatial distribution of surface zooplankton to spatial meso-scale along Magellan Strait.
{"title":"Importance of Remote Sensing for the Study of Spatial Dynamics of Estuarine Neuston from Southern Chile","authors":"J. Cañete, C. Cárdenas, M. Frangópulos, X. Aguilar, J. Díaz-Ochoa","doi":"10.1109/LAGIRS48042.2020.9165571","DOIUrl":"https://doi.org/10.1109/LAGIRS48042.2020.9165571","url":null,"abstract":"Zooplankton aggregation, hydrographic and remote sensing data were employed to relate the spatial dynamics of neustonic communities with chlorophyll a (Chl a) and suspended organic matter (SOM) at a spatial mesoscale (10 to 1000 km) in the southem Chilean fjords system along Magellan Strait, Chile (CIMAR 16: October/November 2010 and CIMAR 25; September/October 2019) in order to identify oceanographic process producing aggregation of neuston. Preliminary evidence of CIMAR 25 shows significant concentrations of Chl a and SOM around Dawson Island (DI), Magellan Strait. During CIMAR 16 important aggregation of specific neustonic taxa (copepodites of Microsetella rosea, larvae of the polychaete Polygordius sp and cyphonautes of the bryozoan Membranipora isabelleana) was observed around DI, Magellan Strait. Satelital images in the area of CIMAR 16 provide evidence of important aggregation of chlorophyll a/SOM around DI. CIMAR Cimar 25 showed that the Chl a and SOM aggregation around DI is recurrent and could to explain the high concentration of neuston around this island to spite of mesotrophic conditions. Remote sensing in this study area provides a tool to understanding oceanographic and topographic factors that potentially regulate the abundance and spatial distribution of surface zooplankton to spatial meso-scale along Magellan Strait.","PeriodicalId":111863,"journal":{"name":"2020 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117042306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-03-01DOI: 10.1109/LAGIRS48042.2020.9165679
M. E. Ferreira, E. B. Silva, F. Malaquias, L. M. S. Teixeira, L. M. Pascoal, N. B. Santos, T. F. Oliveira
In the last decade, the access to geographic information through Web platforms has grown substantially in Brazil, and is now a strategic condition for the social and environmental governance of large biomes such as Cerrado (savanna) and Amazonia. This paper aims to present the first version of the Cerrado Knowledge Platform, designed within the scope of the Critical Ecosystem Partnership Fund (CEPF) for the Brazilian savanna. The Platform aims to provide geospatial and census data, organize and systematize the accumulated knowledge about the Cerrado, and also highlight the actions of researches and social networks in this region. It was developed based on three components: (1) protocols and data formats; (2) adaptation of computational tools for social and environmental analysis and monitoring, in order to notify possible threats to the ecosystem (e.g. burning and deforestation); (3) training and maintenance database component. Although still in a beta version, our platform already has some active features, including access to dynamic land use maps, deforestation, and aerial imagery provided by Unmanned Aerial Vehicles (UAVs). With its enhancement and constant data input from partners, we expect the Cerrado Knowledge Platform can better assist the management of land use and land cover of Cerrado, with a perspective of maintaining key areas for biodiversity conservation.
{"title":"Cerrado Knowledge Platform: A Social And Environmental Management Tool To Conserve Brazilian Savannas","authors":"M. E. Ferreira, E. B. Silva, F. Malaquias, L. M. S. Teixeira, L. M. Pascoal, N. B. Santos, T. F. Oliveira","doi":"10.1109/LAGIRS48042.2020.9165679","DOIUrl":"https://doi.org/10.1109/LAGIRS48042.2020.9165679","url":null,"abstract":"In the last decade, the access to geographic information through Web platforms has grown substantially in Brazil, and is now a strategic condition for the social and environmental governance of large biomes such as Cerrado (savanna) and Amazonia. This paper aims to present the first version of the Cerrado Knowledge Platform, designed within the scope of the Critical Ecosystem Partnership Fund (CEPF) for the Brazilian savanna. The Platform aims to provide geospatial and census data, organize and systematize the accumulated knowledge about the Cerrado, and also highlight the actions of researches and social networks in this region. It was developed based on three components: (1) protocols and data formats; (2) adaptation of computational tools for social and environmental analysis and monitoring, in order to notify possible threats to the ecosystem (e.g. burning and deforestation); (3) training and maintenance database component. Although still in a beta version, our platform already has some active features, including access to dynamic land use maps, deforestation, and aerial imagery provided by Unmanned Aerial Vehicles (UAVs). With its enhancement and constant data input from partners, we expect the Cerrado Knowledge Platform can better assist the management of land use and land cover of Cerrado, with a perspective of maintaining key areas for biodiversity conservation.","PeriodicalId":111863,"journal":{"name":"2020 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128907377","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-03-01DOI: 10.1109/LAGIRS48042.2020.9165625
D. L. Torres, R. Feitosa, L. L. la Rosa, P. Happ, J. Marcato, W. Gonçalves, J. Martins, V. Liesenberg
Knowing the spatial distribution of endangered tree species in a forest ecosystem or forest remnants is a valuable information to support environmental conservation practices. The use of Unmanned Aerial Vehicles (UAVs) offers a suitable alternative for this task, providing very high-resolution images at low costs. In parallel, recent advances in the computer vision field have led to the development of effective deep learning techniques for end-to-end semantic image segmentation. In this scenario, the DeepLabv3+ is well established as the state-of-the-art deep learning method for semantic segmentation tasks. The present paper proposes and assesses the use of DeepLabv3+ for mapping the threatened Dipteryx alata Vogel tree, popularly also known as cumbaru. We also compare two backbone networks for feature extraction in the DeepLabv3+ architecture: the Xception and MobileNetv2. Experiments carried out on a dataset consisting of 225 UAV/RGB images of an urban area in Midwest Brazil demonstrated that DeepLabv3+ was able to achieve in mean overall accuracy and Fl-score above 90%, and IoU above 80%. The experimental analysis also pointed out that the MobileNetv2 backbone overcame its counterpart by a wide margin due to its comparatively simpler architecture in view of the available training data.
{"title":"Semantic Segmentation Of Endangered Tree Species In Brazilian Savanna Using Deeplabv3+ Variants","authors":"D. L. Torres, R. Feitosa, L. L. la Rosa, P. Happ, J. Marcato, W. Gonçalves, J. Martins, V. Liesenberg","doi":"10.1109/LAGIRS48042.2020.9165625","DOIUrl":"https://doi.org/10.1109/LAGIRS48042.2020.9165625","url":null,"abstract":"Knowing the spatial distribution of endangered tree species in a forest ecosystem or forest remnants is a valuable information to support environmental conservation practices. The use of Unmanned Aerial Vehicles (UAVs) offers a suitable alternative for this task, providing very high-resolution images at low costs. In parallel, recent advances in the computer vision field have led to the development of effective deep learning techniques for end-to-end semantic image segmentation. In this scenario, the DeepLabv3+ is well established as the state-of-the-art deep learning method for semantic segmentation tasks. The present paper proposes and assesses the use of DeepLabv3+ for mapping the threatened Dipteryx alata Vogel tree, popularly also known as cumbaru. We also compare two backbone networks for feature extraction in the DeepLabv3+ architecture: the Xception and MobileNetv2. Experiments carried out on a dataset consisting of 225 UAV/RGB images of an urban area in Midwest Brazil demonstrated that DeepLabv3+ was able to achieve in mean overall accuracy and Fl-score above 90%, and IoU above 80%. The experimental analysis also pointed out that the MobileNetv2 backbone overcame its counterpart by a wide margin due to its comparatively simpler architecture in view of the available training data.","PeriodicalId":111863,"journal":{"name":"2020 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS)","volume":"118 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132032681","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-03-01DOI: 10.1109/LAGIRS48042.2020.9285976
C. Almeida, D. Valeriano, L. Maurano, L. Vinhas, L. Fonseca, D. Silva, C. P. Santos, F. Martins, F. C. B. Lara, J. S. Maia, E. R. Profeta, L. O. Santos, F. Santos, V. Ribeiro
(100 - 250 words) Monitoring the conversion of native vegetation has challenged Brazilian government and scientists since the 1980s. In the case of the Amazonian forests, the Amazon Gross Deforestation Monitoring Project - PRODES has developed an effective methodology that provides consistent annual data on deforestation areas on a scale of 1:250,000, since 1988. In this article, we present some aspects of the evolution of this methodology, the key processes to produce accurate deforestation maps during the last 30 years and the new challenges that the Project would face. A central lesson is that no computational technique has, to date, been able to achieve the quality of deforestation maps produced by visual interpretation of satellite images and manual mapping.
{"title":"Deforestation Monitoring in Different Brazilian Biomes: Challenges and Lessons","authors":"C. Almeida, D. Valeriano, L. Maurano, L. Vinhas, L. Fonseca, D. Silva, C. P. Santos, F. Martins, F. C. B. Lara, J. S. Maia, E. R. Profeta, L. O. Santos, F. Santos, V. Ribeiro","doi":"10.1109/LAGIRS48042.2020.9285976","DOIUrl":"https://doi.org/10.1109/LAGIRS48042.2020.9285976","url":null,"abstract":"(100 - 250 words) Monitoring the conversion of native vegetation has challenged Brazilian government and scientists since the 1980s. In the case of the Amazonian forests, the Amazon Gross Deforestation Monitoring Project - PRODES has developed an effective methodology that provides consistent annual data on deforestation areas on a scale of 1:250,000, since 1988. In this article, we present some aspects of the evolution of this methodology, the key processes to produce accurate deforestation maps during the last 30 years and the new challenges that the Project would face. A central lesson is that no computational technique has, to date, been able to achieve the quality of deforestation maps produced by visual interpretation of satellite images and manual mapping.","PeriodicalId":111863,"journal":{"name":"2020 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132202856","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-03-01DOI: 10.1109/LAGIRS48042.2020.9165598
C. L. Mallmann, A. F. Zaninni, W. P. Filho
The biological invasion is considered the second largest global threat to the maintenance and conservation of natural ecosystems biodiversity. Strategies and actions that guide the control and monitoring of invasive species in protected areas are still a challenge on the management of these areas. Remote sensing is potential tool to detect and monitoring these species, gaining a timeline scale and allowing the adoption of more effective control methods. In this study, search to evaluate the vegetation index potential by using multispectral images acquired by UAV as a support on detection and monitoring of invasive plants in Quarta Colônia State Park located on the Brazil’s southern region. A sampling area with a density of invasive plants above 80% was evaluated, with predominance of Psidium guajava and Ligustrum lucidum, generating a large data set from the extracted indexes. Among the evaluated index, the ones that showed the most potential in this study were Green Normalized Difference Vegetation Index (GNDVI), Plant Senescence Reflectance Index (PSRI) and Red Green Ratio Index (RGRI). Believe us that the use of UAVs platforms will be an important tool for the management of invasive species in protected areas.
{"title":"Vegetation Index Based In Unmanned Aerial Vehicle (Uav) To Improve The Management Of Invasive Plants In Protected Areas, Southern Brazil.","authors":"C. L. Mallmann, A. F. Zaninni, W. P. Filho","doi":"10.1109/LAGIRS48042.2020.9165598","DOIUrl":"https://doi.org/10.1109/LAGIRS48042.2020.9165598","url":null,"abstract":"The biological invasion is considered the second largest global threat to the maintenance and conservation of natural ecosystems biodiversity. Strategies and actions that guide the control and monitoring of invasive species in protected areas are still a challenge on the management of these areas. Remote sensing is potential tool to detect and monitoring these species, gaining a timeline scale and allowing the adoption of more effective control methods. In this study, search to evaluate the vegetation index potential by using multispectral images acquired by UAV as a support on detection and monitoring of invasive plants in Quarta Colônia State Park located on the Brazil’s southern region. A sampling area with a density of invasive plants above 80% was evaluated, with predominance of Psidium guajava and Ligustrum lucidum, generating a large data set from the extracted indexes. Among the evaluated index, the ones that showed the most potential in this study were Green Normalized Difference Vegetation Index (GNDVI), Plant Senescence Reflectance Index (PSRI) and Red Green Ratio Index (RGRI). Believe us that the use of UAVs platforms will be an important tool for the management of invasive species in protected areas.","PeriodicalId":111863,"journal":{"name":"2020 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS)","volume":"135 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132604467","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-03-01DOI: 10.1109/LAGIRS48042.2020.9165636
M. B. Nunes, A. Poz, E. Alcântara, M. Curtarelli
Water depth is an important measure for nautical charts. Accurate methods to provide water depth information are expensive and time costing. For this reason, since late 70’s, it started to be estimate by multispectral sensors with empirical models. In the literature there is no investigation using empirical models partitioned in depth intervals, for this reason, we evaluated the accuracy of partitioned and single bathymetric models. The results have shown that to retrieve depth in from 0 to 15 m the single model provided an RMSE of 3.57 m, with a bias of about -0.83 m; while the RMSE for the partitioned model was 2.29 m with a bias of 0.41 m. For updating nautical charts using multispectral sensors it was concluded that the partitioned model can provide a better result than using a single model.
{"title":"Depth Retrieval From A Reservoir Using A Conditional-Based Model","authors":"M. B. Nunes, A. Poz, E. Alcântara, M. Curtarelli","doi":"10.1109/LAGIRS48042.2020.9165636","DOIUrl":"https://doi.org/10.1109/LAGIRS48042.2020.9165636","url":null,"abstract":"Water depth is an important measure for nautical charts. Accurate methods to provide water depth information are expensive and time costing. For this reason, since late 70’s, it started to be estimate by multispectral sensors with empirical models. In the literature there is no investigation using empirical models partitioned in depth intervals, for this reason, we evaluated the accuracy of partitioned and single bathymetric models. The results have shown that to retrieve depth in from 0 to 15 m the single model provided an RMSE of 3.57 m, with a bias of about -0.83 m; while the RMSE for the partitioned model was 2.29 m with a bias of 0.41 m. For updating nautical charts using multispectral sensors it was concluded that the partitioned model can provide a better result than using a single model.","PeriodicalId":111863,"journal":{"name":"2020 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS)","volume":"81 7","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134392566","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-03-01DOI: 10.1109/LAGIRS48042.2020.9165654
L. L. la Rosa, M. Zortea, B. H. Gemignani, Dario Augusto Borges Oliveira, R. Feitosa
Citrus producers need to monitor orchards frequently, and would benefit greatly from having automated tools to analyze aerial images acquired by drones over the plantations. However, analysing large aerial data sets to enable producers to take management decisions that would optimize productivity and sustainability over time and space remains challenging. Motivated by the success of deep learning in computer vision, this work proposes a novel approach based on Fully Convolutional Regression Networks and Multi-Task Learning to detect individual full-grown trees, tree seedlings, and tree gaps in citrus orchards for inventory tracking. We show that the proposal can identify eight-year-old orange trees with accuracy between 95–99% in high-density commercial plantations where adjacent crowns overlap. This quality of detection was achieved on RGB orthomosaics with a pixel size of about 9.5 cm and requires the nominal spacing between adjacent trees as a priori information. Our results also highlight that detecting tree seedlings and tree gaps remains a challenge. For these two categories, classification sensitivity (recall) was between 59–100% and 63–94%, respectively.
{"title":"FCRN-Based Multi-Task Learning for Automatic Citrus Tree Detection From UAV Images","authors":"L. L. la Rosa, M. Zortea, B. H. Gemignani, Dario Augusto Borges Oliveira, R. Feitosa","doi":"10.1109/LAGIRS48042.2020.9165654","DOIUrl":"https://doi.org/10.1109/LAGIRS48042.2020.9165654","url":null,"abstract":"Citrus producers need to monitor orchards frequently, and would benefit greatly from having automated tools to analyze aerial images acquired by drones over the plantations. However, analysing large aerial data sets to enable producers to take management decisions that would optimize productivity and sustainability over time and space remains challenging. Motivated by the success of deep learning in computer vision, this work proposes a novel approach based on Fully Convolutional Regression Networks and Multi-Task Learning to detect individual full-grown trees, tree seedlings, and tree gaps in citrus orchards for inventory tracking. We show that the proposal can identify eight-year-old orange trees with accuracy between 95–99% in high-density commercial plantations where adjacent crowns overlap. This quality of detection was achieved on RGB orthomosaics with a pixel size of about 9.5 cm and requires the nominal spacing between adjacent trees as a priori information. Our results also highlight that detecting tree seedlings and tree gaps remains a challenge. For these two categories, classification sensitivity (recall) was between 59–100% and 63–94%, respectively.","PeriodicalId":111863,"journal":{"name":"2020 IEEE Latin American GRSS & ISPRS Remote Sensing Conference (LAGIRS)","volume":"211 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115054920","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：