Detection of coffee berry necrosis by digital image processing of landsat 8 oli satellite imagery
Coffee berry necrosis is a fungal disease that, at a high level, significantly affects coffee productivity. With the advent of surface mapping satellites, it was possible to obtain information about the spectral signature of the crop on a time scale pertinent to the monitoring and detection of plant...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2020-03-01
|
| Series: | International Journal of Applied Earth Observations and Geoinformation |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S0303243419306853 |
| _version_ | 1818114111711477760 |
|---|---|
| author | Jonathan da Rocha Miranda Marcelo de Carvalho Alves Edson Ampélio Pozza Helon Santos Neto |
| author_facet | Jonathan da Rocha Miranda Marcelo de Carvalho Alves Edson Ampélio Pozza Helon Santos Neto |
| author_sort | Jonathan da Rocha Miranda |
| collection | DOAJ |
| description | Coffee berry necrosis is a fungal disease that, at a high level, significantly affects coffee productivity. With the advent of surface mapping satellites, it was possible to obtain information about the spectral signature of the crop on a time scale pertinent to the monitoring and detection of plant phenological changes. The objective of this paper was to define the best machine learning algorithm that is able to classify the incidence CBN as a function of Landsat 8 OLI images in different atmospheric correction methods. Landsat 8 OLI images were acquired at the dates closest to sampling anthracnose field data at three times corresponding to grain filling period and were submitted to atmospheric corrections by DOS, ATCOR, and 6SV methods. The images classified by the algorithms of machine learning, Random Forest, Multilayer Perceptron and Naive Bayes were tested 30 times in random sampling. Given the overall accuracy of each test, the algorithms were evaluated using the Friedman and Nemenyi tests to identify the statistical difference in the treatments. The obtained results indicated that the overall accuracy and the balanced accuracy index were on an average around 0.55 and 0.45, respectively, for the Naive Bayes and Multilayer Perceptron algorithms in the ATCOR atmospheric correction. According to the Friedman and Nemenyi tests, both algorithms were defined as the best classifiers. These results demonstrate that Landsat 8 OLI images were able to identify an incidence of the coffee berry necrosis by means of machine learning techniques, a fact that cannot be observed by the Pearson correlation. |
| first_indexed | 2024-12-11T03:45:32Z |
| format | Article |
| id | doaj.art-8a0a2927d0c147ca8f5ac72f5aade9b0 |
| institution | Directory Open Access Journal |
| issn | 1569-8432 |
| language | English |
| last_indexed | 2024-12-11T03:45:32Z |
| publishDate | 2020-03-01 |
| publisher | Elsevier |
| record_format | Article |
| series | International Journal of Applied Earth Observations and Geoinformation |
| spelling | doaj.art-8a0a2927d0c147ca8f5ac72f5aade9b02022-12-22T01:22:02ZengElsevierInternational Journal of Applied Earth Observations and Geoinformation1569-84322020-03-0185101983Detection of coffee berry necrosis by digital image processing of landsat 8 oli satellite imageryJonathan da Rocha Miranda0Marcelo de Carvalho Alves1Edson Ampélio Pozza2Helon Santos Neto3Agricultural Engineering Department, Federal University of Lavras, University Campus, P.O.Box 3037, 37200-000, Lavras, Minas Gerais, Brazil; Corresponding author.Department of Agricultural Engineering at the Federal University of Lavras, University Campus, P.O.Box 3037, 37200-000, Lavras, Minas Gerais, BrazilPlant Pathology Department, Federal University of Lavras, University Campus, P.O.Box 3037, 37200-000, Lavras, Minas Gerais, BrazilPlant Pathology Department, Federal University of Lavras, University Campus, P.O.Box 3037, 37200-000, Lavras, Minas Gerais, BrazilCoffee berry necrosis is a fungal disease that, at a high level, significantly affects coffee productivity. With the advent of surface mapping satellites, it was possible to obtain information about the spectral signature of the crop on a time scale pertinent to the monitoring and detection of plant phenological changes. The objective of this paper was to define the best machine learning algorithm that is able to classify the incidence CBN as a function of Landsat 8 OLI images in different atmospheric correction methods. Landsat 8 OLI images were acquired at the dates closest to sampling anthracnose field data at three times corresponding to grain filling period and were submitted to atmospheric corrections by DOS, ATCOR, and 6SV methods. The images classified by the algorithms of machine learning, Random Forest, Multilayer Perceptron and Naive Bayes were tested 30 times in random sampling. Given the overall accuracy of each test, the algorithms were evaluated using the Friedman and Nemenyi tests to identify the statistical difference in the treatments. The obtained results indicated that the overall accuracy and the balanced accuracy index were on an average around 0.55 and 0.45, respectively, for the Naive Bayes and Multilayer Perceptron algorithms in the ATCOR atmospheric correction. According to the Friedman and Nemenyi tests, both algorithms were defined as the best classifiers. These results demonstrate that Landsat 8 OLI images were able to identify an incidence of the coffee berry necrosis by means of machine learning techniques, a fact that cannot be observed by the Pearson correlation.http://www.sciencedirect.com/science/article/pii/S0303243419306853Data miningSpectral behaviorAccuracyColletotrichum ssp.Atmospheric correction |
| spellingShingle | Jonathan da Rocha Miranda Marcelo de Carvalho Alves Edson Ampélio Pozza Helon Santos Neto Detection of coffee berry necrosis by digital image processing of landsat 8 oli satellite imagery International Journal of Applied Earth Observations and Geoinformation Data mining Spectral behavior Accuracy Colletotrichum ssp. Atmospheric correction |
| title | Detection of coffee berry necrosis by digital image processing of landsat 8 oli satellite imagery |
| title_full | Detection of coffee berry necrosis by digital image processing of landsat 8 oli satellite imagery |
| title_fullStr | Detection of coffee berry necrosis by digital image processing of landsat 8 oli satellite imagery |
| title_full_unstemmed | Detection of coffee berry necrosis by digital image processing of landsat 8 oli satellite imagery |
| title_short | Detection of coffee berry necrosis by digital image processing of landsat 8 oli satellite imagery |
| title_sort | detection of coffee berry necrosis by digital image processing of landsat 8 oli satellite imagery |
| topic | Data mining Spectral behavior Accuracy Colletotrichum ssp. Atmospheric correction |
| url | http://www.sciencedirect.com/science/article/pii/S0303243419306853 |
| work_keys_str_mv | AT jonathandarochamiranda detectionofcoffeeberrynecrosisbydigitalimageprocessingoflandsat8olisatelliteimagery AT marcelodecarvalhoalves detectionofcoffeeberrynecrosisbydigitalimageprocessingoflandsat8olisatelliteimagery AT edsonampeliopozza detectionofcoffeeberrynecrosisbydigitalimageprocessingoflandsat8olisatelliteimagery AT helonsantosneto detectionofcoffeeberrynecrosisbydigitalimageprocessingoflandsat8olisatelliteimagery |