Original papersCoffee plantation area recognition in satellite images using Fourier transform
Introduction
Coffee is among the most important agricultural commodities in the world market (Pohlan and Janssens, 2010). It is the second most exported commodity after oil (Gole, 2015). Its popularity and volume of consumption keep growing every year. According FAO (Food and Agriculture Organization of the United Nations) statistics (www.faostat3.fao.org), global coffee production area covered more than 10 million hectares (Pohlan and Janssens, 2010, Gole, 2015) in the world. In order to increase the production yield and improve farming management, coffee plantation has been popularly adopted for widespread commercial sale in the main coffee production countries.
Coffee trees are generally planted in row or line pattern along a specific direction to achieve maximum production yields. In this paper, we present a machine vision scheme to identify row-planted coffee fields in satellite images. Fig. 1(a) shows a satellite image in Kauai, Hawaii. It contains the coffee field, tree forest, buildings and roads in the image. The enlarged image patches of these four categories are presented in Fig. 1(b1)–(e1). The coffee trees display green color and are planted in rows. The forest trees are also green-colored, but grow randomly without any structural arrangement. Due to the variations of terrain, slop and shape of the land, the row patterns of coffee planting could be quite different from region to region. Different growing stages of coffee trees present also complicated texture patterns in satellite images. The satellite images captured in different weather and lighting conditions also cause high variations of coffee plantation regions. The spatial domain-based methods cannot reliably handle the diversity of coffee fields in the satellite image. We thus present a spectral domain method to extract structural features of row-planted coffee fields using the Fourier transform. The Fourier transform allows the extraction of global features of a structural pattern with minimal affection of irregular planting changes and environmental noise.
The row planting of coffee trees shows a line pattern along some specific direction in a region, and the forest trees grow in arbitrary directions without distinct structural arrangement. We thus propose two discriminant features, linear structural ratio and number of rows (density) obtained from the power spectrum of the Fourier transform. Row-planted coffee fields generate high-energy frequency components in a single direction, while naturally-grown forest trees present omnidirectional frequency components in the spectral domain image. The main frequency in the power spectrum indicates the number of parallel lines in a small patch window and, thus, gives the density feature. The density feature for the row-planted coffee filed is equivalent to the number of planting rows in a unit square area, whereas it is only one for the naturally-growing plants. Because the two proposed Fourier features are highly distinctive to separate the row-planted coffee field and irrelevant vegetation area, simple thresholding is robust enough to segment the satellite image. No complicated supervised classification (such as Support Vector Machine, SVM) or unsupervised clustering (such as Fuzzy C-means) is involved for image segmentation. The total coffee planting area and density for mass commercial use can be accurately evaluated from the segmented region and the number of rows in the satellite image. The result can be potentially used for the estimation of the production volume, and eventually the coffee price in the market.
This paper is organized as follows: Section 2 reviews the previous work. Section 3 describes first the material of satellite images used in this study. It then discusses the Fourier transform properties of row-planting patterns in the power spectrum. The structural feature and density feature used for coffee field segmentation are finally defined and described. Section 4 presents the experimental results on a variety of coffee production regions in the world. The effect of changes in the parameter values is also discussed. The paper is concluded in Section 5.
Section snippets
Previous work
Image processing techniques have been used widely for agriculture applications and satellite image analysis. Brosnan and Sun, 2002, Chen et al., 2002 reviewed computer vision for inspection and grading of agricultural and food products. Thorp and Tian (2004) reviewed weed detection techniques in remote sensing images. Meyer and Neto (2008) verified color vegetation indices for crop imaging applications. Ozdogan et al. (2010) reviewed remote sensing of irrigated agriculture. It discussed various
Materials and methods
The images used for this study were obtained from Google Earth Pro. The main image sources are provided by the QuickBird commercial satellite images from DigitalGlobe and LANDSAT-7 satellite images from EarthSat. The test images are retrieved by the geographic coordinates (latitude and longitude) of the target coffee fields. The images of the same location captured in different times are used to evaluate the effect of environmental and lighting changes. In the Google Earth Pro search tool, the
Experimental results
In this section, we present the experimental results on satellite images collected from various coffee plantation areas in the world and in different times. The proposed algorithms in this study are implemented using the C++ programming language and executed on a personal computer equipped with an Intel i7-2600 3.40 GHz processor. The statistics of the structural features are evaluated, and the segmentation results are presented in Section 4.1. The effect of changes in the parameter values on
Conclusions
Plantation agriculture has grown rapidly to increase the production yield and improve farming management. Coffee trees are planted in rows and form a structural texture pattern in the satellite image. To distinguish the coffee field and all other green vegetation areas, the structure instead of color are the only reliable cue to segment the coffee field from the irrelevant areas in the satellite image. In this paper, we have proposed two effective discriminant features to identify the
References (28)
- et al.
Inspection and grading of agricultural and food products by computer vision systems – a review
Comput. Electron. Agric.
(2002) - et al.
Analysis of natural images processing for the extraction of agricultural elements
Image Vis. Comput.
(2010) - et al.
Real-time image processing for crop/weed discrimination in maize fields
Comput. Electron. Agric.
(2011) - et al.
Fast and robust fuzzy c-means clustering algorithms incorporating local information for image segmentation
Pattern Recogn.
(2007) - et al.
An image-processing based algorithm to automatically identify plant disease visual symptoms
Biosyst. Eng.
(2009) - et al.
Machine vision technology for agricultural applications
Comput. Electron. Agric.
(2002) - et al.
Automatic segmentation of relevant textures in agricultural images
Comput. Electron. Agric.
(2011) - et al.
Mapping vineyard leaf with multispectral satellite imagery
Comput. Electron. Agric.
(2003) - et al.
Verification of color vegetation indices for automated crop imaging applications
Comput. Electron. Agric.
(2008) - et al.
Satellite image segmentation with shadowed C-means
Inf. Sci.
(2011)
Automatic detection of crop rows in maize fields with high weeds pressure
Expert Syst. Appl.
Support vector machines in remote sensing: a review
ISPRS J. Photogram. Remote Sens.
Improving plant discrimination in image processing by use of different colour space transformations
Comput. Electron. Agric.
Agriculture satellite image segmentation using a modified artificial Hopfield neural network
Comput. Hum. Behav.
Cited by (17)
Coffee yield prediction using high-resolution satellite imagery and crop nutritional status in Southeast Brazil
2024, Remote Sensing Applications: Society and EnvironmentIntegration of Sentinel optical and radar data for mapping smallholder coffee production systems in Vietnam
2021, Remote Sensing of EnvironmentCitation Excerpt :Most commonly, topographic information is regularly included in many coffee mapping studies (Cordero-Sancho and Sader, 2007; Hebbar et al., 2019; Kelley et al., 2018; Mukashema et al., 2014), which can be attributed to coffee's suitability at higher altitudes (DaMatta et al., 2007). Precipitation or temperature features (Cordero-Sancho and Sader, 2007; Kelley et al., 2018) and texture features (Gaertner, 2017; Gomez et al., 2010; Tsai and Chen, 2017) are rapidly becoming key in difficult-to-map coffee regions. Studies that include texture features often use high-resolution proprietary data such as WorldView (Gaertner, 2017) or Quickbird (Gomez et al., 2010), or focus on mapping or distinguishing other tree crops such as cocoa, rubber, oil palm (Burnett et al., 2019; Gao et al., 2015; Nomura and Mitchard, 2018; Numbisi et al., 2019; Torbick et al., 2016).
A model for Lean and Green integration and monitoring for the coffee sector
2018, Computers and Electronics in AgricultureCitation Excerpt :Nevertheless, the expansion of tropical agricultural commodities, such as coffee, has been considered a major threat to biodiversity (Donald, 2004; Haggar et al., 2017), requiring that practices of optimization of productive processes and reduction of environmental impacts be introduced. With an average annual production of 5.9 Mtons, coffee is one of the most commercialized drinks in the world (Battista et al., 2016) and is the economic backbone of countries throughout Latin America, Asia, and Africa (RAINFOREST, 2017), and its popularity and volume of consumption continues to grow every year (Tsai and Chen, 2017). In Colombia, the production of specialty coffees is extremely important for the national economy, with the livelihoods of about 2 million people dependent on this activity (Cano et al., 2012).
Integration of Unmanned Aerial Vehicle and Multispectral Sensor for Paddy Growth Monitoring Application: A Review
2024, Pertanika Journal of Science and TechnologyCoffee Land Mapping Using Unmanned Aerial Vehicles (UAVs) Multispectral Remote Sensing
2023, Universal Journal of Agricultural Research